Release Notes for MariaDB MaxScale 2.5.21
This page is part of MariaDB's Documentation.
The parent of this page is: Release Notes for MariaDB MaxScale 2.5
Topics on this page:
Overview
MariaDB MaxScale is an advanced database proxy, firewall, and query router.
MariaDB MaxScale 2.5.21 was released on 2022-07-13. This release is of General Availability (GA) maturity.
Notable Changes
The
priority
parameter can now accept a negative value for a server, which eliminates the node from theMaster
selection process for Galera Monitor (galeramon
). (MXS-4165)In earlier releases, the minimum value of the
priority
parameter was0
. When the value was0
, the Galera Monitor (galeramon
) could still select the node asMaster
when no other nodes were available.Starting with this release, when the value of the
priority
parameter is negative, the Galera Monitor (galeramon
) will eliminate the node from theMaster
selection process. When the value is0
, the Galera Monitor (galeramon
) will still select the node asMaster
when no other nodes were available.
Issues Fixed
Can result in a hang or crash
When Binlog Router (
binlogrouter
) is used, a race condition can cause MaxScale to crash. (MXS-4138)When the
connection_keepalive
parameter is0
, a memory leak can cause MaxScale to crash. (MXS-4100, MXS-4141)When
Avrorouter
replicates binary log events that reference , MaxScale can crash. (MXS-4120)
Can result in unexpected behavior
When Schema Router (
schemarouter
) is used, thesys
schema is not automatically ignored for duplicate checks. (MXS-4110)In earlier releases, the
ignore_tables_regex
parameter would have to be used to configure Schema Router (schemarouter
) to ignore thesys
schema when used with MariaDB Enterprise Server 10.6.Starting with this release, Schema Router (
schemarouter
) will automatically ignore thesys
schema for duplicate checks.
When the same query is executed multiple times, the size of the query classifier cache could incorrectly increase, even though the query was only cached once. (MXS-4194)
In earlier releases, repeatedly executing a query could cause the size of the query classifier cache to surpass the value of the
query_classifier_cache_size
parameter, which would cause the query classifier cache to be disabled.Starting with this release, the size of the query classifier cache is updated correctly when the same query is executed multiple times.
When TLS is enabled, MaxScale tries to generate 512-bit RSA keys, which are disabled when OpenSSL is running in FIPS mode. (MXS-4181)
In earlier releases, when this issue occurred on a system using OpenSSL 1.0.1 (such as some older CentOS 7 and RHEL 7 releases), the following error messages would be written to the MaxScale log:
error : (SERVER_NAME); 512-bit RSA key generation failed. error : (SERVER_NAME); Unable to initialize SSL for server 'SERVER_NAME' error : Failed to create a new server.
When
maxctrl show filters
is executed, the output does not contain filter diagnostics. (MXS-4166)When the Cat Router (
cat
) is used, sessions are not always closed, which can trigger a debug assertion to fail. (MXS-4164)The
max_slave_replication_lag
parameter could not be set to a non-zero value with Galera Monitor (galeramon
). (MXS-4160)In earlier releases, when the
max_slave_replication_lag
parameter was set to a non-zero value with Galera Monitor (galeramon
), the following error message would be written to the MaxScale log when Read/Write Split Router (readwritesplit
) tried to route a query to a replica node:Could not find valid server for target type TARGET_SLAVE (COM_QUERY: QUERY_TEXT), closing connection.
When Schema Router (
schemarouter
) is used and duplicate table checks are not disabled with theignore_tables_regex
parameter, the cost of performing the duplicate table check grows very quickly as the number of tables increases. (MXS-4151)When Schema Router (
schemarouter
) is used, query routing performance scales poorly as the number of tables increases. (MXS-4152)When the
connection_keepalive
parameter is non-zero, MaxScale sends pings to servers even if the client is idle. (MXS-4139)When MariaDB Monitor (
mariadbmon
) performs failover, MaxScale writes a containing the replication user's user name and password in plain text to the MaxScale Log. (MXS-4115)Starting with this release, the user name and password are masked in the statement written to the MaxScale Log.
When multiple servers are selected as
Master
with Read Connection Router (readconnroute
), therank
parameter is ignored for all servers except the last one. (MXS-4132)In earlier releases, this could impact query routing in some configurations. For example, let's say that two servers are selected as
Master
: the first hasrank=primary
, and the second hasrank=secondary
. Whenrouter_options=master
is set, Read Connection Router (readconnroute
) chooses the second server, because the rank of the first server is ignored.
When a service uses the
targets
parameter to route queries to a second service, the second service is not shown as a target in the output ofmaxctrl list services
. (MXS-4127)If the value of the
query_classifier_cache_size
parameter is set high, executingmaxctrl show qc_cache
can cause MaxCtrl to run out of memory. (MXS-4121)
Related to install and upgrade
When the MaxScale package is installed, the
/etc/maxscale.cnf.d/
directory is not automatically created. (MXS-4134)
Platforms
In alignment to the MariaDB Corporation Engineering Policy, MariaDB MaxScale 2.5.21 is provided for:
CentOS 7 (x86_
64) Debian 9 (x86_
64, ARM64) Debian 10 (x86_
64, ARM64) Debian 11 (x86_
64, ARM64) Red Hat Enterprise Linux 7 (x86_
64) Red Hat Enterprise Linux 8 (x86_
64, ARM64) Rocky Linux 8 (x86_
64, ARM64) SUSE Linux Enterprise Server 15 (x86_
64, ARM64) Ubuntu 18.04 (x86_
64, ARM64) Ubuntu 20.04 (x86_
64, ARM64)