Detected different number in database and cache. Triggering panic on node

[Simon G_]

Hi!

I have an old instance of Confluence (6.3.1). I know this version is clearly outdated, and we are aware that we need to replace it. We're working on that, but for now, this old instance remains quite useful to us.

For the past few days, I’ve regularly found Confluence down, displaying the following error:

Non-Clustered Confluence: Database is being updated by another Confluence instance. Please see http://confluence.atlassian.com/x/mwiyCg for more details.

Looking at the logs, I see the problem always occurs after this error:

2025-03-28 13:01:01,104 WARN [Caesium-1-4] [confluence.cluster.safety.DefaultClusterSafetyManager] onNumbersAreDifferent Detected different number in database [ -910347570 ] and cache [ -822303782 ]. Cache number last set by [ not clustered ]. Triggering panic on node [ not clustered ]

I simply restart Confluence to resolve the issue, but it keeps happening. Is there anything I can try to work around this problem? For example, resetting the cache, repairing database tables... or even disabling the task causing this error (which might be bugged since it's registering negative numbers)?

Many thanks in advance.

[Shawn Doyle - ReleaseTEAM]

Hi @Simon G_ 

If you haven't read this already, this doc may help.

https://confluence.atlassian.com/confkb/confluence-cluster-issues-180847200.html

Also, the troubleshooting guide: 

https://confluence.atlassian.com/display/CONF54/Cluster+Troubleshooting

[Simon G_]

Thank you for your feedback.

The first link provides interesting insights, even though I'm not using a cluster but a standalone Confluence instance..

Among the possible causes mentioned in this article, the only one I can't completely rule out is this one:

Your database is taking a long time to commit (e.g. it is doing a back up that lasts over 30 secs and freezes commits). In this case the cluster safety job sends a new value to the database, but when it checks next time in 30secs, the previous commit hasn't occurred so it still fetches the previous value which is out of sync with its cached value.

Moreover, I’ve noticed that the issue occurs almost every time on the hour (1:00, 13:00, etc.) and is always preceded by a series of errors:

Caused by: com.mysql.jdbc.exceptions.jdbc4.CommunicationsException: Communications link failure

I should probably investigate this further...

Unfortunately, second link doesn't give so much information in my case as I'm not using a cluster and link to solution "'Database is being updated by an instance which is not part of the current cluster' Error Message" appears to be broken (just like http://confluence.atlassian.com/x/mwiyCg)