Disabled Discourse Whos Online Plugin for Root Cause Analysis

Recent Action

Meta Discourse has a warning that the whos online plugin is "experimental" and does not work well on large forums. We are not large relatively speaking; but this seems like the logical best place to start to search for the root cause of our recent 502 HTTP errors, since I noticed the DB hangs during that error.

Screen Shot 2020-09-15 at 4.14.49 PM

Background

Recently, we have seen issues with the Discourse PostgreSQL DB "hanging" which causes a HTTP 502 error and so as a part of our root cause analysis, I have disabled this plugin for the time being.

The Meta Discourse support site does not run this plugin either, so that might be a clue as well.
After disabling this plugin today, the site seems faster, but that might be an illusion.

The Discourse Whos Online plugin is also not an official Discourse plugin (maybe it is "officially supported" now, not sure?), so it is "technically" not supported (please correct me if this is wrong); so for now, it's disabled... let's see how it goes and hope the 502 errors disappear.

It's kinda' a shame, because many people, including me, like that plugin; but maybe it turns out it is not the problem :slight_smile:

Frankly, I thought for a few seconds about a simple crontab that restarts the data container every so many hours; but I really prefer to find out what is causing the DB "hanging" problem versus just burying the problem with restart and never getting to the root cause.

Thank you for understanding and your patience.

See Also:

2 Likes

Yes, each user displaying information of each user --> it goes exponentially!
And yes, a stupid crontab job is, well, stupid.

1 Like

Yes, totally agree.

Automatic "restart the dead or hung container jobs" are really sloppy and poor sys admin work!

I am fully behind root cause analysis!

Let's see if the problem goes away with this plugin disabled.

:slight_smile:

So far, only subjectively speaking, the site seems must faster and reliable with the Whos Online plugin turned off.

I have not see any errors or "hangups" so far after disabling and all the DB errors in the logs, which appeared before, are gone (so far).

Let's see what happens when the site is stormed by a bot swarm next time it happens.

2 Likes

Removed this Whos Online plugin completely from Discourse configuration.

See also:

Update:

This site has not seen any problems at all since we disabled and removed the "Whos Online" Discourse plugin.

Therefore, I consider this plugin the "root cause" and this matter closed; and we will not use that plugin on our site.