I’ve been running 12.0.4 for a while, when I had to do some maintenance on the filestore. I stopped apache and unmounted the nfs filestore. Then remount store, restart apache. This procedure might have caused the following problem, maybe not:
Accessing files has become very slow. I tested uploading a 300kb file thorougly, and found that it will appear on the server file system within <0.5 seconds, with loglevel=0 file creation is logged, but the request won’t return until another 15 seconds have elapsed.
I restarted apache and redis, disabled brute force, emptying lock and bruteforce tables, upgrade to 12.0.5, no improvement.
Any hint what may cause this 15 second delay (sometimes only 9 seconds)? No fancy file handling apps installed which might interfere, AFAICS.
You should take some time to tell us your whole setup (software & hardware) and what kind of maintenance was that.
It’s Debian9 with Apache2, virtualized on decent hardware. memcache.local=APCu and memcache.locking=Redis and LDAP auth.
Maintenance was firewall a upgrade that interrupts nfs traffic so the nfs store needed to be unmounted.
From my observation, there must be some wait timeout in nextcloud code, after storing the file is done and logged, but I don’t see any LDAP or other connection attempts during that nap time.
Some days ago, the server was fast again after a reboot. Today (now) the same happening again. Very weird: A file upload via web interface is taking 20 seconds (logging of PUT in apache access.log), but 15 seconds before the request is logged complete by apache the nextcloud client already notifies about the successful download.
Investigating further: there’s no unexpected tcp traffic, but on localhost the traffic to redis is a whopping 276000 packets between port 6379 and one single port, i.e. for a single file put. Unfortunately, disabling any memcaching doesn’t help, so the redis traffic is just a symptom.