GitHub user ingox added a comment to the discussion: OOBM Issue during KVM HA

The root cause is HAProxy TCP timeouts (50 seconds) applied to the MariaDB VIP. 
CloudStack keeps long-lived pooled DB connections; HAProxy closes idle TCP 
sessions after 50 seconds, causing stale/broken pooled connections. Eventually 
CloudStack’s DB pool becomes exhausted and throws _Timeout waiting for idle 
object_.

Update your proxy config and reload the haproxy:
_frontend mysql_frontend
  bind 192.168.30.16:3306
  mode tcp
  option tcplog
  timeout client  1h
  default_backend galera_backend_

_backend galera_backend
  mode tcp
  balance roundrobin
  option tcpka
  timeout server  1h
  timeout connect 5s
  option mysql-check user haproxy
  default-server fall 3 rise 2 inter 2000
  server prod-node01 192.168.30.2:3307 check
  server prod-node02 192.168.30.11:3307 check
  server prod-node03 192.168.30.15:3307 check_

Update your /etc/cloudstack/management/db.properties to:
_db.pool.testOnBorrow=true
db.pool.validationQuery=SELECT 1
db.pool.testWhileIdle=true_

GitHub link: 
https://github.com/apache/cloudstack/discussions/12353#discussioncomment-15424627

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to