Video Storage deletion issues

[dgossage]

I have a system with 15 cameras.

It's running Centos7 and the storage is using zfs raidz6 4TB WD Red drives. @15TB useable space currently

I'm starting to run into the issue where I am writing faster than the system is cleaning up the images and directories.
I've even taken to writing a cleanup script to use perl to unlink files faster than the filter will with rm.

Curious if anyone else has come across the issue. ZFS does have a little more overhead when removing files so I have considered if I should just move back to xfs. I have also thought maybe I should split zoneminder between multiple servers spreading the storage out.

I'm looking at adding more cameras in future so trying to plan out a sustainable model going forward.

[marcopete87]

I have a system with 15 cameras.

It's running Centos7 and the storage is using zfs raidz6 4TB WD Red drives. @15TB useable space currently

I'm starting to run into the issue where I am writing faster than the system is cleaning up the images and directories.
I've even taken to writing a cleanup script to use perl to unlink files faster than the filter will with rm.

Curious if anyone else has come across the issue. ZFS does have a little more overhead when removing files so I have considered if I should just move back to xfs. I have also thought maybe I should split zoneminder between multiple servers spreading the storage out.

I'm looking at adding more cameras in future so trying to plan out a sustainable model going forward.

I still wonder which kind of issue do you have.

[dgossage]

Mainly that the system can't delete the older files faster than the new ones are written.

Curious if anyone else has systems with that many cameras running to one storage system with any similar problems.

[iconnor]

It gets better with h264 storage, as you are only deleting 1 or 2 files instead of thousands.

[dgossage]

That's coming in 1.31 if i recall?

[marcopete87]

It gets better with h264 storage, as you are only deleting 1 or 2 files instead of thousands.

Iconnor, as you told me some time ago, raid5 doesn't boost performance on write; does raid6 do the same?

dgossage, how many seconds last an event? I had same issue (with buffer overflow due disk slowness), mitigated by setting 60 seconds events.

[dgossage]

I have it set to default I think so they are in 10 minute chunks. So yes I could likely change it to 1 minute a piece which would give me more directories but with fewer images in each. So hopefully each rm command would return faster.

I doubt raid6 is great for performance but with 4TB drives rebuilding one when a drive dies takes a bit of time so I like the extra parity.

[iconnor]

raid6 is definitely a better thing than raid5, but offers no write performance improvements.

What would be nice would be a raid with a lazy redundancy. Because I wouldn't really care if I lost a file or two, As long as most were there.

I think zfs and btrfs have some more interesting raid options, but I havn't had time to look into them.