Huge load on maintenance task related to GP, following upgrade NC19 -> NC20, forcing to disable GP

[pierfu]

Hi,
a few days after upgrading Nextcloud from 19 to 20.0.7, we had to face a critical issue on a production server, ands it seamed directly related to GroupFolders.

• huge lines in NC log, like ~5.000.000 characters, of failed requests, with :
  1/ recurring pattern : An exception occurred while executing 'DELETE FROM "oc_filecache_extended" WHERE "fileid" IN (?...?) ' with params [id1,... idN] :\n\nSQLSTATE[HY000]: General error: 7 number of parameters must be between 0 and 65535", ,"class":"Doctrine\DBAL\Driver\PDOException"},"DELETE FROM "oc_filecache_extended" WHERE "fileid" IN (?...?) ", [19017890,19537170,19459948,18911007,15822686,"And 89981 more entries
  2/ GroupFolders appears many times in these huge log lines : {"file":"/var/www/nextcloud/nextcloud-20.0.7/apps/groupfolders/lib/ACL/ACLManager.php","line":75, "function":"getRootStorageId","class":"OCA\GroupFolders\ACL\ACLManager","type":"->","args":[]}

• problem was solved first times wirth "php occ maintenance:repair", that seamed to kill the job responsible for blocking everything, but never in a long-term way

• the only long-term workaroud was to disable groupfolders completely

So,

• has anyone known similar issue with after upgrade to Nextcloud 20 ?
• could this be a maintenance issue with GroupFolders, which would require too much work to Nextcloud Core at the same time, instead of asking this maintenance with a maximum entries set ?

I join an excerpt of the log.
tail_-n1_nextcloud_abbr.log

[brtptrs]

We are seeing similar huge query statements, and many long running mysql requests that often include group-folders.
NC 20.0.7
I can not tell if the problem started with NC20. But we noticed problems recently, that wight coincide with the upgrade from 19 to 20.

[pierfu]

Hi,
any update on this big issue ?

[ASerbinski]

That duplicate is not a duplicate. It involves php unnecessarily/inefficiently processing a long list of files in the trashcan.

[pierfu]

Just to be clear, when i pubished this issue, by saying maintenance issue with GroupFolders i did not mean maintenance mode at all, but maintenance task run by cron or whatever.

[pierreozoux]

@pierfu do you still have issues? Maybe an update fixed it?

[pierfu]

@pierfu do you still have issues? Maybe an update fixed it?

I got no answer on this issue, and it happened on a big production instance, so the only thing i'm sure about is that i won't take that risk to just try and mybe break everything. Sorry :/

Is this extension planned on beeing merged in Nextcloud core ? I can see on https://apps.nextcloud.com/apps/groupfolders that Nextcloud GMBH is now editor. That could maybe help on maintaining this app more efficiently.

[pierreozoux]

What do you mean nextcloud core?
It is part of the supported app:
https://docs.nextcloud.com/server/latest/admin_manual/installation/apps_supported.html

So if you have support with Nextcloud GMBH, they will gladly help you.

[pierfu]

Thanks for your answer.
By core I mean not-an-app.

I missed this supported-apps page, and this possibility of asking support to Nextcloud GMBH.
But from what i undersand in pricing presentation, the only way of having quick support garanty (quick or not, what does mean : "Reaction time: 1h, 4h, 12h or 24h" ??), is Premium plan, but i definitly can't afford "100 users for €95.50/user/year" cost.
And your suggestion means that you can't rely on this Github issues, not really.

So i can't just try again GP and see if it crashes.
The really big problem with this app is that it changes data management too deeply : one cannot just try it, make GP shares, and disable it, because shares are not managed as they are in the core way.
I spent hours with my client on re-creating shares that were managed by GP, especially link shares.

[pierreozoux]

I understand.
You could "duplicate" this instance in an other env to try.
And yes, the "community" support is for Home users mainly, it is not recommended to run big production instance without support or knowing what you are doing.
(But i'm not representing Nc here, I'm just part of the community, not an employee, and what I'm saying comes from my experience of Nextcloud communication)

[pierfu]

Thanks Pierre,
we already have a staging instance, mainly for validating NC updates, but we never experienced nor could reproduce the issue on staging, surely because of scale and load difference.
So for now, i honestly can't recommend to re-enable GP, mainly because :

• i had no technical answer on the issue
• GP changes NC too deeply to be easy to enable/disable.

I think we can close even if not resolved.

[pierreozoux]

There is a somehow related issue here:
#1215

As you want feel free to close or not, your call :)
(I think it is good to keep this issue open, it is documenting the current state of this app, but this is only my opinion :) )

[pierfu]

OK, il leave it opened.

[solracsf]

This was mainly fixed by 57cf151
Queries are now made in chunks to avoid that General error: 7 number of parameters must be between 0 and 65535