I have a huge problem with my owncloud which might is caused by a misconfiguration of apache.
Since we have a user with a huge data dir (he uploaded 75GB), the owncloud crashes now every 1-2 hours. The snyc client gets an "socket operation timeout", webusers see a white page or a server error and our monitoring gives the following:
agent name: cloud
Datum/Uhrzeit: 17-4-2015 08:30:01 (GMT + 3)
Current Working set size: 1533.05 MB / Max value: 1024
Current Page file usage: 58187.38 MB / Max value: 15360
Summarized you can say, that owncloud seems to stop operating on the server side after a while when huge sync loads are proceeded on the sync client over a longer period of time (some hours) - even if the server system itself seems to continue operating normal. Just a quick reboot puts the server back to normal operation - until a critcal amout of sync load had been processed again.
Let me give some facts on the situation to give an impression of what's going on:
oc8 runs virtualized on a single ubuntu 14.04 LTS server with 4 cores on a new Dell PowerEdge R730xd, 4GB on a 20GB hdd with about 50% free space left. storage of the data dir is mounted into linux from a Windows network share. CPU and RAM are operating in the lowest 10% of their given server ressources.
oc8 is performing fine when operating in regular daily business (normal sync loads - about 2-3GB a day by 50 users)
performance problems always seems to occurs when users are uploading/downloading huge amounts of personal data (>5-10GB), in a stronly nested folder hierarchy, at a time using the owncloud sync client for Windows (even in version 1.8).
The sync client gets disconnected from the server after a while with a "socket operation timeout" error. When trying to login to the owncloud server, you might get a blackscreen (hang) OR you're facing network HTTP problems concerning the apace/oc8, although the performance key indicators (RAM, SWAP, HDD, CPU) seem to be within acceptable ranges. system parameters are unsuspicious at any time (much free RAM, much free SWAP, etc) - except HTTP response times on the apache port of oc8 which exceed 1500ms by far in case of the fault which is described here.
Does anyone have an idea what this can be and how it can be solved?