Released 1.36.29 The Memory Remains

[iconnor]

#Changes since 1.36.28

- update web/ajax.log.php to contents from master. Fixes errors causing log view to not work. Fixes #3606
- use ajax() instead of getJSON so that we can specify no timeouts.. This prevents log queries from stacking up overloading the db
- Check for definition of CAMBOZOLA defines. The purpose is just to ease running the 1.36 UI against a 1.37 database.
- Added option ZM_AUTH_CASE_INSENSITIVE_USERNAMES to match mixed case Usernames to lower case usernames in database #3516
- Move LIBAVCODEC_VERSION_CHECK so that it is defined when the include files are under ffmpeg. Maybe fixes build with 5.1.2?
- Test for matches[operator]. Fixes #3607

1.36.29
Full Changelog

[pat2]

Thanks Isaac. It seems not yet loaded on the repo... https://launchpad.net/~iconnor/+archive ... inder-1.36

Isaac, I (it's Michele) asked your friendship on Linkedn

[ZMBrian]

1.36.29 not loaded on repo as of Oct 12, 15:36 UTC. I have installed 1.36.28 and can't start cameras and can't access log file from the browser.

[iconnor]

Packages should be available now

[ZMBrian]

Confirmed update to 1.36.29. However, I still can't reactivate my cameras.

When I select the camera's "Function" from "None" to "Mocord," the browser page goes blank.

The log shows the following errors:

"10/12/22, 3:42:57 PM PDT",web_php,,1206,FAT,"View 'function' does not exist",/usr/share/zoneminder/www/index.php,293
"10/12/22, 3:42:57 PM PDT",web_php,,1206,ERR,"actions can no longer be performed without POST.",/usr/share/zoneminder/www/index.php,209

[iconnor]

Use the main monitor edit instead of the function popup for now

[ZMBrian]

Use the main monitor edit instead of the function popup for now

That suggestion worked, thank you.

[sideband]

I installed the update from 1.36.28 to .29 then rebooted the Ubuntu Server 22.04 VM.

Within 5 minutes, the VM crashed. By crashed I mean it was completely unresponsive except for pings (it would respond to pings, but couldn't SSH in, or get any response to input in the hypervisor's VM console).

I killed the VM, and re-started it, this time logging in on the console immediately and running "top". Top showed memory use climb rapidly, then swap use climbing rapidly until full. OOMkiller couldn't keep up.

Luckily I was able to restore from last night's backup (only losing DB entries since the backup... I'll clean up the filesystem later) and the cams are back up and running, but, previous memory issues (from .24 onward) seem to be amplified in .29.

Maybe I'm an edge case... not sure, but thank goodness for backups. Any ideas on what could cause memory to fill up without anything touching shm? (shm remained at 1%/400some MB used during the memory use's meteoric climb to a crash state)

[dougmccrary]

No vm, but it does seem a bit hard on memory.
If it’s crashing I haven’t noticed. 22.04

[iconnor]

There's not much in there that would eat more ram than before.

Which process is consuming ram?

[sideband]

Presently one of the zmc processes is holding 13GB res and 15GB virt.

This is on .28 after restoring the VM. I'm not really wanting to reinstall .29 presently, but it was the zmc processes on .29, as well.

[iconnor]

And which version didn't consume ram?

[sideband]

Last one I remember not going ape on RAM usage was before .24

[sideband]

The memory usage on .28 as of this morning. I'm still wary of attempting to upgrade to .29 again because of that memory issue.
To get here, I had to restart individual monitors individually. The monitor which was holding more than half the VMs RAM is behaving itself - for now.