Firstly, I seem to have this problem (*) too, although I’m using the out-of-the-box FreeBSD ports install, which uses MySQL and mod-php (5.6) on Apache 2.4, in mpm-worker configuration. Nextcloud is 13.0.2. FreeBSD is 12-CURRENT. System has 32G DRAM, 4-disk RAIDZ, 50G Swap.
(*) This problem being that httpd instances grow in size, seemingly without bound. Ultimately swap fills and shortly after the system crashes.
Now this seems really odd to me, because there ought to be no mechanism for an httpd instance to persist for long enough to exhibit a memory leak, right?
I wouldn’t say that I have large quantities of small files, either. Just documents and the like, for two users.
I do share contacts and calendar with clients on a couple of mac, iphone and android devices.
So: how is the discovery that lots of small files are implicated count as a “solution”? Still growing without bounds for me, and I have to keep an eye on it and restart apache every so often to avoid system instability and crashing.
This is a fairly new phenomenon. I don’t recall it happening under earlier versions of nextcloud. I can’t point to any particular change happening due to configuration changes, but there are many moving parts, and they all change and upgrade over time. If anyone has any suggestions about what I might do to tame the beastie, I’d appreciate it. I restarted apache half way through writing this message, and the four running instances of httpd are consuming a little less than 100M/minute…