Aerospike Server Community Edition 5.0.0.33

Description

• Before upgrading to 5.0.0.4 or later, customers already running 5.0.0.3 in a cluster which is an active XDR destination (i.e. an XDR destination also acting as a source XDR or that may in the future act as a source XDR), may want to consider stopping all XDR traffic to the cluster. (Enterprise Only)
  • In the unlikely case of an upgraded 5.0.0.4 or newer node going down while 5.0.0.3 nodes are still running, those nodes would mistakenly ship XDR writes even if forward has not been explicitly enabled. 
  • Enterprise Licensees may contact Aerospike Support for any questions before upgrading.
• Cross-Datacenter Replication (XDR) running in Aerospike Server version 5.0 can only connect to XDR destination clusters running versions 4.1 or newer.
• All nodes must be upgraded to Aerospike Server version 4.9.0.7 or newer. Please upgrade to the latest 4.9.0.x release prior to upgrading to Aerospike Server version 5.0.
  • Aerospike Server version 5.0 will not form a cluster with version 4.8 nodes or earlier.
  • When upgrading the Aerospike Server refer to the Special Upgrades documentation.
• The Aerospike Server will not start if obsolete configuration parameters are set in the Aerospike configuration file (aerospike.conf).
  • Refer to: Configuration parameters that must be removed for Aerospike Server version 5.0
• ATTENTION: As of Aerospike Server version 4.9, expirations and evictions are disabled by default.
  • For expirations, nsup-period must be configured (new default is 0).
  • For evictions, the relevant high-water-disk-pct/high-water-memory-pct/mounts-high-water-pct must be configured (new defaults are 0).
  • For details refer to the Namespace Retention Configuration documentation.
• CentOS Linux 8 End of Life is scheduled for December 2021.
  • For more information please refer to: FAQ - CentOS Project shifts focus to CentOS Stream

New Features

Improvements

Bug Fixes

• Enterprise & Community
  • [AER-6412] - (QUERY) If a query transaction setup fails in certain ways, improper cleanup may cause a crash.
• Enterprise Only
  • [AER-6413] - (SECURITY) Small memory leak during LDAP login process.
  • [AER-6411] - (XDR) Race when dynamically adding a DC under write load may cause a crash.

Known Issues

• Enterprise & Community
  • [AER-6340] - (KVS) HLL string or blob elements sized 31 to 254 bytes may be treated differently depending on msgpack standard. Fixed in version 5.4.0.1.
    • Different client versions use 2 different msgpack standards. If HLLs have string or blob elements in the size range 31-254 inclusive and are accessed via clients using differing msgpack standards, a given element can be added to an HLL as 2 different elements. Also, even if individual HLLs were built consistently, operations combining them will give incorrect results if they were built using different standards. Once any of this has happened, it is not possible to fix the affected HLL(s).
  • [AER-6444] - (CDT) For putItems(), the policy check is skipped for empty ordered maps. Fixed in version 5.0.0.37.
  • [AER-6438] - (QUERY) A background operations query may become inaccessible via the job monitor before it completes. Fixed in version 5.0.0.36.
• Enterprise Only
  • [AER-6451] - (KVS) Race condition between sprig element counters and changing sprig roots when rebuilding index during warm restart may cause a crash. Fixed in version 5.0.0.38.
  • [AER-6427] - (KVS) For data-in-index namespaces, durable deletes will cause an unnecessary assertion on a warm restart. Fixed in version 5.0.0.35.
  • [AER-6423] - (LOGGING) For strong-consistency namespaces, non-roster nodes are not always logged when rebalancing. Fixed in version 5.0.0.34.

Updates