Dashboard fails due to S3 timeout


#1

My client has recently run into problems that look similar to those in this earlier thread (now closed):

I’m seeing similar timeout issues on the dashboard, and they go away if I do SiteSetting.backup_location = 'local' from the rails console, where it was previously set tos3.

The number of files in the s3 bucket isn’t especially large though, at about 600. IT’s not shared with uploads from the site.

From the thread above, it looks like work was being done on this issue, but the thread was closed with no report on progress. What was the result?

The problem arose when doing updates from within the site admin area. Would I expect to see a different result if i rebuild the container from the outside?


(Gerhard Schlager) #3

It hasn’t been fixed yet. Out of curiosity, are all 600 files backups?


#4

Yes, they’re all backup files. It seems likely we can get a temporary fix by manually deleting some. However that would mean the problem would recur down the track.


#5

We’ve reduced the s3 bucket to 57 files (1 month of daily files, plus monthly ones before that). We still can’t load the dashboard. It doesn’t look like the file count is the whole picture here though.

I’ve been considering implementing a script to regularly prune the old backups, but it seems like it would be much better for discourse to manage the whole of the backup life-cycle. It appears that it wouldn’t help with the current situation anyway.


#6

So it seems likely that the problem is with S3, but apparently not with the number of files. Unless I need to reset some sort of cached state?

Is this affecting other people?

Any ideas how to proceed?


(Gerhard Schlager) #7

I’m working on a fix. I’ll let you know when it’s done.


#8

Thanks for letting me know.