really slow ZFS on 5.3-12, works fine on 5.3-11!

[rhradec]

I uploaded to 5.3-12 last Friday, and after expend the weekend trying to find what was going on with my zfs pools crawling to a halt, varying from 0.5 to 2MB/sec (they're usually go from 100 to 250mb/sec)!

Today I remembered that I had updated last friday, and try to reboot on 5.3-11... to my surprise, my zfs pools are back to 100-250mb/sec!!!

booting on 5.3-12 again, they are back to "slow motion mode".

I have no other conclusion but that 5.3-12 have some weird problem with ZFS.

So, in case someone is having trouble with zfs disk speed, try to boot on 5.3-11!!

It was a huge scare over the weekend, because I thought all my pools where failing... But thank god it was just the update!!

weird thing is that there was NO error messages... nothing!! just a huge slow down...

my 2 cents!

 

[dcsapak]

what exactly do you boot? i guess you mean the kernel (which do have different versions)

can you post 'pveversion -v' of both slow and fast boots ?

 

[Binary Bandit]

Interesting. We're having similar issues where a cluster node slows to the point where the watchdog triggers a hard reboot. This all started after an upgrade to 5.3-12.

I've moved the most heavily loaded node back to 5.3-11 and it's been stable for three days now.

We're using Ceph though ...

Will follow this thread ... here's my thread: https://forum.proxmox.com/threads/cluster-nodes-unstable.53136/

James