ZM1.29 on CentOS7...

[rockandroller]

ZM1.29 on CentOS7... "oops".

It's a dual-quad Xeon3Ghz with 18GB RAM

Still testing, now have seven IP cams hooked up through an 8-port PoE switch.
Activated all seven cams for "MoCord" and immediately caught a couple of events each - and system load went up to "overdrive". All eight cores were running 100%

Set all seven cams back down to "monitor" and now the system is chugging away with load average "1" (of possible 8...) and 3.5 GB ram in use. I'm wondering what it is that I might have terribly misconfigured... any hints?

Still getting "error while decoding frame" (every 400 frames) on all five HIKvision cams, (but not on either D-link cam) - but I'm not sure if that's relevant to this CPU load issue or not...

PS: ZM system is set to "medium bandwidth" defaults

[bbunge]

Set the resolution on all the cams down low (320x240 or whatever your cameras do) and the frame rate to 5 FPS (at the cameras not in Zoneminder). With 18 GIG RAM it is unlikely you are running out of tmpfs (/dev/shm) space but a "df" will show you what you have used. You may want to run mysqltuner to see what is going on with mariadb server. I had to set my innodb_buffer_pool_size = 2G (i can actually go back to 1G if I need to with 65K events in the database). And make sure you do not have any trailing spaces in any settings you change.

bb

[rockandroller]

Will try that!

Reading the notes on /dev/shm in another recent thread... How do I know how big "memory files" should be?

/dev/shm is 8 gigs, with 2.5GB used at present. All these tempfs were created by default... I made just one single 2TB partition for the CentOS install, with the intention that ZM would not run out of room - if for some reason it was writing some stuff to a smaller "OS" partition.

Code: Select all

Filesystem               Size  Used Avail Use% Mounted on
/dev/mapper/centos-root  2.0T   10G  2.0T   1% /
devtmpfs                 8.8G     0  8.8G   0% /dev
tmpfs                    8.8G  2.4G  6.5G  27% /dev/shm
tmpfs                    8.8G  8.9M  8.8G   1% /run
tmpfs                    8.8G     0  8.8G   0% /sys/fs/cgroup
/dev/cciss/c0d0p2        497M  294M  204M  60% /boot
/dev/mapper/centos-home   87G   91M   87G   1% /home
tmpfs                    1.8G  4.0K  1.8G   1% /run/user/994
tmpfs                    1.8G     0  1.8G   0% /run/user/0

[rockandroller]

OK, with the streams tuned down as low as they go (320x176 for the two D-links, and 1280x720 for the five HIKs - all at 4FPS) - the CPU load is much lower... now running about 0.5- 0.9 when there is some action, and "idling" at about 0.2

However, that is with all monitors set to Modect, which is what I should have had them on previously anyway ( Been a couple months since I played with this, and on first try had accidentally set them to Mocord, which is not needed for this application...)

But what I notice ( maybe its odd, maybe not?) is that there are two /usr/bin/zm -m processes for each of cam1&2 ( the dlinks, which are capturing their stream 1 in default JPEG mode) and three /usr/bin/zm -m for cams 3-7 ( the HIKs, which are set to h264 mode for their stream1).

Is that indeed normal?

And BTW, is there any "best practice" camera mode?

[rockandroller]

Just to answer my own question, it seems with the HIKvision DS-2CD2035-I on Centos7 - the best practice seems to be "Use FFmpeg".
( it solves all the 'decoding frame' errors I was getting) * see viewtopic.php?f=34&t=24586

At any rate, all seems to be working good - 7 cams doing fullHD@4FPS and the server doesn't break a sweat

And not one single error message in the log [insert giant thumbs up icon in the general direction of the ZoneMinder devs]...