Confluence Server Restart Causes Notification Emails For Old Page Edits

[Paul]

In our Confluence 6.1.3 Data Centre Setup we have noticed that a server restart causes email notifications to be sent for page edits and likes that occurred several weeks ago.

I can't find a bug for this. Has anybody else noticed this problem?

thanks

Paul.

[AnnWorley · Accepted solution]

Hi Paul,

Your question brings up more questions, I hope you don't mind letting me know:

• If you check your email queue on the Confluence Admin>Email Queue page, do you see emails backed up in there?
• Do people receive emails from Confluence as expected between server restarts?
• When you mention a server restart, does that mean one node or all?
• How many nodes are in the Data Center instance?
• Are other scheduled jobs such as flushing the index queue running as expected?

There is a condition in Confluence where a job gets stuck and all scheduled jobs stop running. Scheduled jobs may stall and fail to process if one job becomes stuck

One theory is, if you are running into this ^ a restart would start all the jobs again, possibly sending out old emails that were queued.

I look forward to hearing more about the issue. 

Thanks,

Ann

[Paul]

Many thanks for the reply Ann,

There are no emails backed up at the moment but I didnt check this before the server restart

Yes users do receive notifications, the restart seemed to cause a resend of old notifications.

We have 2 Confluence Nodes in our Data Centre Instance. We only restarted one node after which some users reported receiving notifications relating to old page edits etc.

Looking at the scheduled jobs, on node 2, a number of the scheduled jobs show the next executing as December 2017 (last year)  :-

• Check Cluster Safety
• Clean Temporary Directory
• Cluster Cache Compaction
• Conversion Queue Monitor
• Deferred File Deletion
• EhCache Cache Compaction
• Flush Local Task Queue
• Flush Mail Queue
• Flush Task Queue

Node 1 has next execution dates in the future for all jobs. I am worried now that there is some underlying problem with node 2. Our current plan is to bring up another node, if its scheduled jobs look ok on that  (next execution dates are in the future) then we will rebuild node 2.

[Paul]

Our new node 3 has correct (future) execution  dates for the scheduled jobs. We spotted that node 2 had the wrong locale (en_us rather than en_uk) but not sure if this is the cause of the problem.

I guess we wont know for certain until next time we restart one of the live nodes.