[
https://issues.apache.org/jira/browse/KARAF-7353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17475425#comment-17475425
]
Jean-Baptiste Onofré commented on KARAF-7353:
---------------------------------------------
So, you are using the HA lock. Can you please share the
system.properties/config.properties to see your config (especially the
runlevel). It's just to cleanly reproduce your issue.
> JDBC Lock Long Running Transaction and PostgreSQL Vacuum
> --------------------------------------------------------
>
> Key: KARAF-7353
> URL: https://issues.apache.org/jira/browse/KARAF-7353
> Project: Karaf
> Issue Type: Bug
> Components: karaf
> Affects Versions: 4.3.5
> Environment: Karaf 4.3.2 (same code as 4.3.5 and master as far as
> PostgreSQLJDBCLock is concerned)
> PostgreSQL 10
> Reporter: Benjamin BONNET
> Assignee: Jean-Baptiste Onofré
> Priority: Critical
>
> Hi,
>
> Karaf H/A relies on long running transactions (see PostgreSQLJDBCLockJDBCLock
> implementation), each Karaf instance running its own transaction to keep
> (master) or try to grab (slaves) the lock. The presence of those transactions
> prevents PostgreSQL vacuum to do its jobs correctly, and vacuum job generates
> those logs :
> {{WARNING: oldest xmin is far in the past
> HINT: Close open transactions soon to avoid wraparound problems.}}
> So, letting our Karaf instances run will lead to a TXID exhaustion.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
