nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Jira email notification very occasionally fails; how to troubleshoot?

Kelly Schoenhofen

We've had a few verified instances (verified as much as I can, really) where an issue was assigned to someone and they didn't get an email notification. After the first couple reports, and all the logs appeared clean, and the users are swearing Jira's emails aren't in their spam folder or anything like that, I turned on the "Outgoing mail log" (but not at the debug level).

Another report came in this morning; I checked the latest atlassian-jira-outgoing-mail.log and the last entry was written about 20 hours ago, the day prior.

Any ideas on how I can troubleshoot this further? Notifications work 99% of the time, it's this 1% that is being noticed however.

-Kelly

Accepted answers

&(*&)#)_*#@@(*)(@*)(*@

maybe there is something with the data intregity and some factory/clauseHandler/ContextHandler is failing sillently , i doubt than the MailImpl would fail without a noticable stacktrace ..

because the timing of the deadlocks seem to correspond with the email failures.

database deadlocks ? .. ok..

You should definitely get some warning/error/info in the log (from the datasource, and from jira) ..and if such thing is happening you can expecte some transactions rollback , without detailled info and acces to your log files , there is little hope to find help.. ask your local java/database stacktrace reader or request support from atlassian

All comments

AhmadDanial

Hey there, Kelly.

I would recommend you to look into JIRA's Notification Troubleshooting guide here to understand the possibility of the notifications are not being sent properly:

https://confluence.atlassian.com/display/JIRAKB/Notifications+Troubleshooting

Check if the users have the Notify Me enabled in their user profile, do the users belong in the respective Permission Scheme, and also check the Mail Queue if there are any pending mails that are not flushed properly.

Warm regards,

Danial

Kelly Schoenhofen

Ah, figured it out. It is only logging WARN and ERROR's.

Will enabling debug logging have it log every email sent?

Kelly Schoenhofen

All good tips

It's been a psuedo-random set of users (no pattern that I can discern), and they get 99% of the rest of their notifications, just the occasional one only when they are assigned an issue. That's the only email that we've seen fail. And it works 99% of the time.

So they have Notify Me enabled (because they get other emails from Jira), and they belong in the respective permission scheme - they are users in the project, and the current assignee.

Mail queue is empty. I've never seen it have anything in it.

Kelly Schoenhofen

Two (or three) questions - with outgoing email logging enabled, shouldn't it have logged that it sent the email?

The outgoing log has logged messages (like this:

2013-06-07 15:31:05,055 ERROR [Jira incoming mail] Sending mailitem To='XXX' Subject='XXX'

But it doesn't for days at a time - why? Why does it log some emails but not others?

Also, why does it say [Jira incoming mail] on it?

&(*&)#)_*#@@(*)(@*)(*@

you may only need to check the mail queue in the adminstration section

1 : the user may not recieve the notification because the "notify my action" setting is not set in their profile.

2 : the mail may end in a span folder , check the mail queue for the mail to be send

3 : check the mai error queue

4: you may have a postfunction failing only for a certain user ? do you have some groovy postfunction ?

when a grrovy or whatever postfunction is failing ie : throwing an exception , all remaining posfunction are discarded , and without an event , no notification

5 : worst case , there is something in your custom templates/mail/created_issue.vm that crashs for that particular user , same as for 4 , check for null and exception.

best regards

Kelly Schoenhofen

I had already looked into all five of your suggestions (good suggestions though!) and found nothing. We are getting deadlocks occasionally though with ms-sql in spite of doing all the documented changes for Jira & MS-SQL. I'm working with support in a ticket for this and they are having me making some other changes to the DB. The reason I bring this up

is because the timing of the deadlocks seem to correspond with the email failures. The downside is this may take awhile to validate - it's hard getting people to notice they didn't get 1 email out of 100, and then be positive the 1 email they didn't get wasn't because of an email filter, rule, or a second device they used to read and delete the email (and then forgot they did so).

&(*&)#)_*#@@(*)(@*)(*@

maybe there is something with the data intregity and some factory/clauseHandler/ContextHandler is failing sillently , i doubt than the MailImpl would fail without a noticable stacktrace ..

because the timing of the deadlocks seem to correspond with the email failures.

database deadlocks ? .. ok..

Kelly Schoenhofen

Atlassian support worked with me and determined that the deadlocks were caused by MS-SQL having issues with the read_committed_snapshot not being set correctly for Jira.

Ultimately, this is what I did:

ALTER DATABASE jira_db_name
SET READ_COMMITTED_SNAPSHOT ON
WITH ROLLBACK IMMEDIATE;

Then:

SELECT sd.is_read_committed_snapshot_on
FROM sys.databases AS sd
WHERE sd.[name] = 'jira_db_name';

now returns a "1" (prior it was 0).

Since that moment, our deadlocks ceased. It's been 21 days and I haven't gotten any reports of email problems. So it doesn't mean things are perfect, but without evidence of further email shenanigans, I'm marking this as fixed. Thanks.

Welcome!

It looks like you're new here. Sign in or register to get started.

Quick Links

Find more posts tagged with

jira

jira-administration