I also saw this without the error message. I clicked Comment, and no comment posted/no error. I refreshed, and the comment was still cached in the the comment box.
@meks@clenneville Unfortunately that's the error that we serve regardless of the problem. Did you by chance see what the/if there was an error response in your network requests? If not, mind taking a look if it happens again?
In the meantime, I'll look and see if we're seeing an escalation in errors on the discussion endpoint.
@thomasrandolph handled the fix. If I'm not mistaken, we particularly fixed the case where the comment was a command only [1]. Seems like the situation described in this issue is when you have just comments or comments mixed with quick actions, maybe? Might be worth investigating the error handling on those cases.
While I don't think I encountered the exact problem in this issue, I had very clear reproducible situations where errors occurred and we never got any report of them. In my testing, it was a command that errored + some other comment text. In that case, the comment would submit and the commands that errored would disappear into the ether.
It seems like this issue is the opposite problem: mentioning someone + some comment text. The mentioning is failing (maybe?) and nothing is happening, but the submission failure/error is being reported.
Probably also worth noting this gem from that thread:
In any other scenario, it will continue to show the more generic error from before.
We don't really get any info from the backend about what happened, but as noted:
were the API to start sending back [robust] errors when [some error occurs], we would need to add front end code for that.
@kencjohnston do you remember if you applied any quick action with that comment? From the system notes, there doesn't seem to have been any activity, but just checking.
do you remember if you applied any quick action with that comment?
I don't remember. I will say that if you suspect the bug is related to something like quick actions that my specific comment didn't have that I'm only 40% confident that is the correct comment.
I had the same problems that commenting/submitting issues returned an error "... check your network connection and try again" in my gitlab-ee 13.9.6 installation.
It turned out the the elasticsearch did not start properly. After fixing the elasticsearch issue, the problem disappeared.
Why would submitting a comment require elasticsearch? This dependency does not make sense.
Super cool you were able to pinpoint it. Do we need to keep these two separate issues?
Are you able to identify similar timeouts to see how frequently this is happening and whether the severity/priory need to be increased? (I note this issue is 2 while the other is 3)?