Aerospike Server FE Release Notes
-
Release Date: March 16, 2023
Known Issues
Updates
-
Release Date: March 8, 2023
Known Issues
Updates
-
Release Date: January 24, 2023
Known Issues
Updates
-
Release Date: December 8, 2022
Known Issues
Updates
-
Release Date: December 2, 2022
Known Issues
Updates
-
Release Date: November 17, 2022
Known Issues
Updates
-
Release Date: March 16, 2023
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6617] - (DEPLOYMENT) The Aerospike server may fail to start in containers with cgroup v1 when the available CPUs are a proper subset of the actual CPUs.
- [AER-6618] - (EXPRESSIONS) Deleting a bin via a write operation expression will cause a crash.
- [AER-6615] - (SMD) A module that is continually retransmitting may block another module's 'set' operations.
- Enterprise & Federal Only
- [AER-6619] - (BALANCE) For
strong-consistency
namespaces, a stale appeal acknowledgement may cause a crash.
- [AER-6616] - (XDR) Removing connector seed nodes while a DC is disconnected may cause an assertion on reconnection.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: March 8, 2023
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6612] - (CDT) Writing an ordered map with a map index bigger than the content may corrupt the map.
- [AER-6614] - (CDT) With result type
EXISTS
, a GET_ALL_BY_KEY_LIST
operation on an ordered map always returns false.
- [AER-6610] - (DEPLOYMENT) The Aerospike server may fail to start in some containers when the available CPUs are a proper subset of the actual CPUs.
- [AER-6613] - (KVS) For
single-bin
, data-in-memory
namespaces, writes using the delete-record operation may free the bin data twice and cause a crash.
- [AER-6611] - (SINDEX) Dropping a secondary index on which there are active queries may cause an assertion or other bad symptoms.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: January 24, 2023
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6601] - (CLUSTERING) When a node fails to cluster due to a cluster size limit, it leaks memory.
- [AER-6597] - (QUERY) A secondary index query may return incorrect results if the bin type changes from CDT to non-CDT during the query.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: December 15, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Improvements
- Enterprise, Federal & Community
- [TOOLS-2286] - (ASADM) Bundle asadm with Python 3.10.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: December 8, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6583] - (INFO) A
sindex-exists
command where the 'ns' parameter does not match a namespace on the node will cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: December 2, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Improvements
- Enterprise, Federal & Community
- [AER-6581] - (LOGGING) Changed logging level for 'starting/finished aggregation query job...' messages from info to debug.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6580] - (SINDEX) Deleting an indexed list or map containing multiple elements with the same value may leave inaccessible keys in the secondary index.
- Enterprise & Federal Only
- [AER-6582] - (XDR) Setting DC context configuration item
tls-name
after adding seed nodes which require it may cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: November 11, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6576] - (SINDEX) Race between garbage collection and update of a secondary index entry may result in the erroneous removal of the entry.
- Enterprise & Federal Only
- [AER-6572] - (XDR) Not setting DC context configuration item
auth-user
when auth-mode
requires it will cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: October 3, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Improvements
- Enterprise & Federal Only
- [AER-6570] - (DEPLOYMENT) With
index-type flash
namespaces, allow the server to start as a non-root user if the flash-index-friendly '/proc/sys' values have been set.
Bug Fixes
- Enterprise & Federal Only
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: September 14, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6568] - (QUERY) For single-bin namespaces, primary index queries erroneously fail without starting.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: August 30, 2022
- For upgrading, or the unlikely event of downgrading, refer to Special Upgrades.
- Special upgrade instructions for Aerospike Server 6.1.
- Before you upgrade, find any secondary index names that exceed the new 64-character limit (including a null terminator), delete them, then recreate the sindex with a shorter name after the upgrade.
- When upgrading, the secondary index
sindex-create
applies to the entire namespace, if the set parameter is omitted.
- Downgrade from Aerospike Database 6.1 to 6.0.
- If you have to downgrade from a deployment of server 6.1 that has a secondary index stored in shared memory, delete the following shared memory segments after stopping the server (asd), in order to reclaim this memory.
- Downgrading from Server 6.1 will force a cold start.
- Server 6.1 removes
num-partitions
and adds the sindex-stage-size
configuration parameter for secondary indexes in the namespace context. sindex-stage-size
is similar to the existing primary index index-stage-size
configuration parameter. If the sindex-stage-size
value is changed across a restart, any secondary indexes for the namespace will be rebuilt.
- There is a possibility of partial de-duplicate (repeated records) in secondary index query (SI query) responses for SI queries that use list or map type secondary indexes. Any record that satisfies such a query at least once will be included, but it is possible and normal to see a record more than once.
- Attempting to create a secondary index which already exists now returns success/OK instead of an error.
- List Operation
EXISTS
return type has been added.
- Map Operation
EXISTS
return type has been added.
Map indexing
and List indexing
can be created at any depth. For example, an index on a map inside of a bin that is inside of another map can be created.
- Server 6.1 includes a simple feature key file allowing only for a single-node cluster. The feature key will not be valid for versions after Server 6.1.
- Please be aware of your package manager's handling of existing configuration files during installation. If the configuration file was updated we keep the existing *.deb or *.rpm respective configuration file. It is the uninstall of the package where these systems diverge.
- Debian and Ubuntu (*.deb) will leave the modified configurations in place.
- Redhat (*.rpm) will rename the configuration file with an *.rpmsave extension.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
- Debian 9 Stretch LTS has reached its End of Life on June 30 2022. We will drop support for Debian 9 in the upcoming Server 6.2 release.
New Features
- Enterprise, Federal & Community
- [AER-6545] - (SINDEX) Added ability to define secondary indexes on elements within CDTs (Collection Data Types), by using CDT
contexts
.
- Enterprise & Federal Only
- [AER-6543] - (SINDEX) Added support for warm restart of secondary indexes.
Improvements
- Enterprise, Federal & Community
- [AER-6530] - (BATCH) Added an option to ensure that a sub-transaction error that stops a response will be included in the response.
- [AER-6532] - (CDT) Added
EXISTS
return type, which may be used to optimize certain 'GET' operations.
- [AER-6535] - (INFO) Added 'sindex-name' field to
query-show
output.
- [AER-6548] - (INFO) Added
smd-show
command to retrieve raw SMD module data.
- [AER-6531] - (LOGGING) Added 'long-queries' to
in-progress
ticker line, and renamed queries_active to long_queries_active
.
- [AER-6537] - (QUERY) Partially de-duplicate responses to range queries on list or map secondary indexes.
- [AER-6546] - (SINDEX) When indexing list or map keys or values, iterate in msgpack directly instead of converting to as_list or as_map.
- [AER-6550] - (SINDEX) A secondary index without a set name now applies to the entire namespace.
- [AER-6551] - (SINDEX) The size limit for a secondary index name has been reduced to 64.
- [AER-6547] - (STATS) Added
entries_per_bval
and entries_per_rec
secondary index cardinality statistics.
- [AER-6534] - (UDF) New as_map implementation improves (de-)serialization of ordered maps when executing UDFs.
- Enterprise & Federal Only
- [AER-6549] - (DEPLOYMENT) Automatically install an Enterprise evaluation feature key.
- [AER-6538] - (SINDEX) For
index-type
flash namespaces, use smaller internal batch sizes for garbage collection and queries, to minimize time under lock.
- [AER-6540] - (STATS) Added
bytes_shipped
namespace statistic in XDR.
- [AER-6544] - (XDR) For connector destinations, use digest affinity (instead of round-robin) to ensure that a record always goes to the same node.
- [AER-6565] - (XDR) Distribute recovery/rewind requests across service threads using partition affinity, avoiding lock contention and significantly improving performance.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: March 16, 2023
- This General Availability release 6.0.0.14 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6617] - (DEPLOYMENT) The Aerospike server may fail to start in containers with cgroup v1 when the available CPUs are a proper subset of the actual CPUs.
- [AER-6618] - (EXPRESSIONS) Deleting a bin via a write operation expression will cause a crash.
- [AER-6615] - (SMD) A module that is continually retransmitting may block another module's 'set' operations.
- Enterprise & Federal Only
- [AER-6619] - (BALANCE) For
strong-consistency
namespaces, a stale appeal acknowledgement may cause a crash.
- [AER-6616] - (XDR) Removing connector seed nodes while a DC is disconnected may cause an assertion on reconnection.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: March 8, 2023
- This General Availability release 6.0.0.13 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6612] - (CDT) Writing an ordered map with a map index bigger than the content may corrupt the map.
- [AER-6610] - (DEPLOYMENT) The Aerospike server may fail to start in some containers when the available CPUs are a proper subset of the actual CPUs.
- [AER-6613] - (KVS) For
single-bin
, data-in-memory
namespaces, writes using the delete-record operation may free the bin data twice and cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: January 24, 2023
- This General Availability release 6.0.0.11 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6601] - (CLUSTERING) When a node fails to cluster due to a cluster size limit, it leaks memory.
- [AER-6597] - (QUERY) A secondary index query may return incorrect results if the bin type changes from CDT to non-CDT during the query.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: December 15, 2022
- This General Availability release 6.0.0.10 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Improvements
- Enterprise, Federal & Community
- [TOOLS-2286] - (ASADM) Bundle asadm with Python 3.10.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: December 2, 2022
- This General Availability release 6.0.0.9 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Improvements
- Enterprise, Federal & Community
- [AER-6581] - (LOGGING) Changed logging level for 'starting/finished aggregation query job...' messages from info to debug.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6580] - (SINDEX) Deleting an indexed list or map containing multiple elements with the same value may leave inaccessible keys in the secondary index.
- Enterprise & Federal Only
- [AER-6582] - (XDR) Setting DC context configuration item
tls-name
after adding seed nodes which require it may cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: November 11, 2022
- This General Availability release 6.0.0.8 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6576] - (SINDEX) Race between garbage collection and update of a secondary index entry may result in the erroneous removal of the entry.
- Enterprise & Federal Only
- [AER-6572] - (XDR) Not setting DC context configuration item
auth-user
when auth-mode
requires it will cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: October 3, 2022
- This General Availability release 6.0.0.7 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Improvements
- Enterprise & Federal Only
- [AER-6570] - (DEPLOYMENT) With
index-type flash
namespaces, allow the server to start as a non-root user if the flash-index-friendly '/proc/sys' values have been set.
Bug Fixes
- Enterprise & Federal Only
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: September 14, 2022
- This General Availability release 6.0.0.6 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6568] - (QUERY) For single-bin namespaces, primary index queries erroneously fail without starting.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: August 23, 2022
- This General Availability release 6.0.0.5 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6566] - (SINDEX) When cold starting storage-backed
data-in-memory
namespaces, finding a later but expired version of a secondary-indexed record will assert.
- Enterprise & Federal Only
- [AER-6567] - (SINDEX) For
index-type flash
, a client non-durably deleting a secondary-indexed record may cause an assertion.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: August 17, 2022
- This release is superseded by 6.0.0.5.
- This General Availability release 6.0.0.4 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Improvements
- Enterprise & Federal Only
- [AER-6562] - (SINDEX) For
index-type flash
, adjust secondary index(es) directly on non-durable deletes, instead of relying on garbage collection. [Note: this introduced an issue fixed in AER-6567.]
Bug Fixes
- Enterprise, Federal & Community
- [AER-6564] - (CDT) A context operation on a CDT with top-level list which changes the CDT size such that it crosses a 256, 64K, or 1M byte boundary may cause a crash.
- [AER-6563] - (EXPRESSIONS) For namespaces with data not in memory, a client-provided object used as a bin in a call may cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
-
Release Date: August 4, 2022
- This General Availability release 6.0.0.3 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6555] - (UDF) For namespaces with data not in memory, a UDF that writes multiple bins and would exceed
write-block-size
may cause a crash.
- Enterprise & Federal Only
- [AER-6553] - (XDR) A source node may never discover all destination nodes if it starts with partial visibility of the remote cluster.
- [AER-6554] - (MIGRATION) For
strong-consistency
namespaces, a failed appeal may block a node from starting migrations.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: June 30, 2022
- This General Availability release 6.0.0.2 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6541] - (EXPRESSIONS) A cond in an expression fails to set the return type, which may allow an illegal expression to build and cause an assertion.
- [AER-6533] - (LOGGING) Set index memory is not included in 'total-bytes' and 'used-pct' on the
memory-usage
ticker line.
- Enterprise & Federal Only
- [AER-6536] - (LOGGING) TLS I/O error may log incorrect errno.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
-
Release Date: May 4, 2022
- This General Availability release 6.0.0.1 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6529] - (SINDEX) Garbage collection does not decrement
memory_used_sindex_bytes
, which may cause erroneous stop-writes or eviction.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
-
Release Date: April 27, 2022
- This General Availability release 6.0.0.0 contains all hotfixes deployed during the release candidate cycle.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
New Features
- Enterprise, Federal & Community
- [AER-6511] - (PACKAGING) Added support for Debian 11.
- [AER-6512] - (PACKAGING) Removed support for Debian 8.
- [AER-6500] - (QUERY) Enhanced secondary index queries: query by partition, paginate, throttle, apply quotas.
Improvements
- Enterprise, Federal & Community
- [AER-6488] - (BALANCE) For AP namespaces, made namespace context configuration item
replication-factor
dynamic.
- [AER-6489] - (BALANCE) Removed service context configuration item
paxos-single-replica-limit
.
- [AER-6499] - (BATCH) Added support for batch writes, deletes, and UDFs.
- [AER-6509] - (DEPLOYMENT) With
auto-pin numa
, handle client interfaces with differing numbers of queues.
- [AER-6492] - (EXPRESSIONS) Removed support for predicate expressions.
- [AER-6485] - (KVS) Fail duplicate resolutions if they are overwhelming the write queue.
- [AER-6504] - (KVS) Check 'rw-hash' and 'proxy-hash' for timed out transactions more frequently, to help the client recycle sockets more efficiently.
- [AER-6510] - (MIGRATION) Do not retransmit updated or deleted records.
- [AER-6515] - (PACKAGING) Ensure that the 'aerospike' user has a secure nologin shell.
- [AER-6506] - (SCAN) Unified scans with new queries.
- [AER-6491] - (STATS) Latency histograms now include time spent processing other service epoll events that precede the transaction request.
- [AER-6507] - (STATS) Base
heap_efficiency_pct
on active rather than mapped memory.
- [AER-6513] - (STORAGE) Detect and mitigate records which are only partially written during an unclean shutdown.
- [AER-6514] - (STORAGE) Verify that a device not previously used as an Aerospike storage device was erased before use.
- [AER-6503] - (UDF) Removed unused 'gen' field from
udf-get
response, and added sanity check of 'content-len' field in udf-put
command.
- Enterprise & Federal Only
- [AER-6495] - (ACL) Added a separate 'truncate' privilege (no longer under 'write'), 'udf-admin' privilege, and 'sindex-admin' privilege.
- [AER-6487] - (BALANCE) To restore availability, revived nodes now require all nodes to be trusted (i.e. no e-flags).
- [AER-6505] - (LOGGING) Separated relevant XDR per-DC ticker line items into per-namespace lines.
- [AER-6501] - (SECURITY) Added support for a special Enterprise build which is FIPS 140-2 compliant.
- [AER-6483] - (XDR) Added DC namespace context configuration item
compression-threshold
to specify a size threshold for compression.
- [AER-6490] - (XDR) Added value 'no-bins' to DC namespace context configuration item
bin-policy
, to ship record metadata only, including keys, to connectors.
- [AER-6497] - (XDR) Reapply set shipping filters, and other configured filters, during retries.
- [AER-6498] - (XDR) Changed automatic throttling to reduce throughput more quickly, with a lower floor.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: April 20, 2022
- This is a Developer Preview Release.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
Bug Fixes
- Enterprise, Federal & Community
- [AER-6526] - (SINDEX) Truncating a namespace or set with a secondary index, and more than 2 billion records per node, may overflow a garbage collection queue and crash.
- Enterprise & Federal Only
- [AER-6527] - (XDR) Dynamically removing seed nodes while trying to connect may cause connections to be skipped or cause a crash.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates
-
Release Date: April 11, 2022
- This is a Developer Preview Release.
- When upgrading to server version 6.0, you must either trim or wipe (zeroize) the Aerospike header on the devices. A modified record storage format was introduced by AER-6513, which adds protection against partially written records during an unclean shutdown.
- For upgrading, or the unlikely event of downgrading, please refer to Special Upgrades.
- BREAKING CHANGE: The
truncate
privilege is now a separate privilege and is no longer under write
.
- The batch operations capability has been made comprehensive. We now fully support batch reads, writes, deletes, and UDF commands.
- The previously deprecated Predicate Expression API has been removed. Use Aerospike Filter Expressions (added in version 5.2).
- Service context configuration
paxos-single-replica-limit
has been removed.
- Configurations, Metrics, and Server Logs related to Scans are now referred to as Primary Index Queries (pi queries). For example, scan_basic_complete is now referred to as pi_query_long_basic_complete.
- The
jobs
info command is now officially deprecated and will continue to be supported through a deprecation period, after which it will be removed. jobs
are now redundant with the new secondary index building methods and the following new info commands:
query-show
replaces query-list
and the deprecated scan-show
command.
query-abort
replaces query-kill
and the deprecated scan-abort
command.
- The
batch-without-digests
configuration parameter has been removed.
- A number of
Best Practices
will be checked at startup. The Best Practices can be enforced by enabling the new enforce-best-practices
configuration item.
- Several configuration parameters have been renamed or removed. Please refer to
Config, Metrics, and Info Changes in Aerospike Database 6.0
.
- Use Prometheus Exporter version 1.2.1 or later to run Prometheus Exporter with this version of Aerospike Server.
New Features
- Enterprise, Federal & Community
- [AER-6511] - (PACKAGING) Added support for Debian 11.
- [AER-6512] - (PACKAGING) Removed support for Debian 8.
- [AER-6500] - (QUERY) Enhanced secondary index queries: query by partition, paginate, throttle, apply quotas.
Improvements
- Enterprise, Federal & Community
- [AER-6488] - (BALANCE) For AP namespaces, made namespace context configuration item
replication-factor
dynamic.
- [AER-6489] - (BALANCE) Removed service context configuration item
paxos-single-replica-limit
.
- [AER-6499] - (BATCH) Added support for batch writes, deletes, and UDFs.
- [AER-6509] - (DEPLOYMENT) With
auto-pin numa
, handle client interfaces with differing numbers of queues.
- [AER-6492] - (EXPRESSIONS) Removed support for predicate expressions.
- [AER-6485] - (KVS) Fail duplicate resolutions if they are overwhelming the write queue.
- [AER-6504] - (KVS) Check 'rw-hash' and 'proxy-hash' for timed out transactions more frequently, to help the client recycle sockets more efficiently.
- [AER-6510] - (MIGRATION) Do not retransmit updated or deleted records.
- [AER-6515] - (PACKAGING) Ensure that the 'aerospike' user has a secure nologin shell.
- [AER-6506] - (SCAN) Unified scans with new queries.
- [AER-6491] - (STATS) Latency histograms now include time spent processing other service epoll events that precede the transaction request.
- [AER-6507] - (STATS) Base
heap_efficiency_pct
on active rather than mapped memory.
- [AER-6513] - (STORAGE) Detect and mitigate records which are only partially written during an unclean shutdown.
- [AER-6514] - (STORAGE) Verify that a device not previously used as an Aerospike storage device was erased before use.
- [AER-6503] - (UDF) Removed unused 'gen' field from
udf-get
response, and added sanity check of 'content-len' field in udf-put
command.
- Enterprise & Federal Only
- [AER-6495] - (ACL) Added a separate 'truncate' privilege (no longer under 'write'), 'udf-admin' privilege, and 'sindex-admin' privilege.
- [AER-6487] - (BALANCE) To restore availability, revived nodes now require all nodes to be trusted (i.e. no e-flags).
- [AER-6505] - (LOGGING) Separated relevant XDR per-DC ticker line items into per-namespace lines.
- [AER-6501] - (SECURITY) Added support for a special Enterprise build which is FIPS 140-2 compliant.
- [AER-6483] - (XDR) Added DC namespace context configuration item
compression-threshold
to specify a size threshold for compression.
- [AER-6490] - (XDR) Added value 'no-bins' to DC namespace context configuration item
bin-policy
, to ship record metadata only, including keys, to connectors.
- [AER-6497] - (XDR) Reapply set shipping filters, and other configured filters, during retries.
- [AER-6498] - (XDR) Changed automatic throttling to reduce throughput more quickly, with a lower floor.
Known Issues
- If you use ‘equality’ secondary index queries that return a small number of records, you may see a latency increase in server releases 6.0-6.2. A partial fix is included in server 6.3, and a complete fix for server 6.4.
- If your secondary index query only returns 0 records or 1 record, meaning that it is acting as a unique index, you should consider converting such a query to a get/exists operation by creating a self-managed reverse index (lookup table).
- Spark Connector version 3.5 starts using batch writes when possible, a feature which was introduced in Aerospike server version 6.0. Consult the Spark Connector rate limiting documentation before you upgrade the server. Pay attention to throttling write throughput, and use Spark Connector 3.5.5 or later, which fixes throttling behavior to conform to the
aerospike.transaction.rate
.
- Known Issues: Server
Updates