[#MXS-3654] Critical read error with MX on sync forced mariadb cluster

I'm using latest maxscale on Debian 10.10 with 10.5.11-MariaDB with a read-write split router configuration.

The mariadb galera cluster nodes have syncing forced enabled with this parameter:
wsrep_sync_wait=7

That means consistency is enabled on the cluster at the cost of performance, so that if something is written at moment t, when a user reads at moment t+1 on another node, this node has to wait to be synced with all t-time writesets before answering.

Nevertheless, a critical read error happens every day in the process of inserting an item and then reading it, using Maxscale's read-write split router.

The read part fails as if the nodes of the cluster were not synced as it is not finding the just inserted row.

[maxscale]

threads=auto

#log_notice=1

#log_debug=1

#log_info=1

syslog=false

auth_connect_timeout=3s

auth_read_timeout=3s

auth_write_timeout=3s

[MySQLMonitorLocal]

type=monitor

module=galeramon

servers=marianode1,marianode2,marianode3

user=maxscale

password=hidden

monitor_interval=1000

backend_connect_timeout=3s

backend_write_timeout=3s

backend_read_timeout=3s

# =============== Services ================ #

[RW2]

type=service

router=readwritesplit

# deprecated

#localhost_match_wildcard_host=1

servers=marianode1,marianode2,marianode3

max_slave_connections=2

user=maxscale

password=hidden

max_connections=350

strip_db_esc=true

# ===============  Listeners ================ #

[RWListenerLocal]

type=listener

service=RW2

protocol=MariaDBClient

port=3310

address=127.0.0.1

[RWListenerLocalSocket]

type=listener

service=RW2

protocol=MariaDBClient

socket=/var/lib/maxscale/rwsplit.sock

# =============== Servers ================ #

[marianode1]

type=server

address=10.100.20.50

port=3306

protocol=MariaDBBackend

#persistpoolmax=100

#persistmaxtime=60s

#rank=primary

[marianode2]

type=server

address=10.100.20.51

port=3306

protocol=MariaDBBackend

#persistpoolmax=100

#persistmaxtime=60s

#rank=secondary

[marianode3]

type=server

address=10.100.20.52

port=3306

protocol=MariaDBBackend

#persistpoolmax=100

#persistmaxtime=60s

#rank=primary

[MXS-3654] Critical read error with MX on sync forced mariadb cluster Created: 2021-07-01 Updated: 2021-08-15 Resolved: 2021-08-15
Status:	Closed
Project:	MariaDB MaxScale
Component/s:	N/A
Affects Version/s:	2.5.13
Fix Version/s:	N/A

[MXS-3654] Critical read error with MX on sync forced mariadb cluster Created: 2021-07-01 Updated: 2021-08-15 Resolved: 2021-08-15