In '/etc/zfs/zed.d/zed.rc' the option 'ereport.fs.zfs.data' is disabled, if you activate it, then error messages might be delivered. On Proxmox the 'pvestatd' daemon queries the storages every second, the disks shouldn't go into sleep.
Shouldn't the events also be visible when executing 'zpool events'? They are missing there too. Also I'm expecting 'ereport.fs.zfs.io' errors, not 'ereport.fs.zfs.data' errors. The affected disks can go to sleep as they are not directly used as storage for PVE. They contain datasets which are bindmounted into some LXCs.
Then it will not get send, as far as I know, the setting is only for those in zpool events. Only as a work around, you could disable the sleep with some udev rules.
Yeah, I think ZFS 0.7.2 changed something, with 0.6.9 the events were still generated. I think we can mark this issue as resolved, ZED is working as it should.
Don't mean to hijack your thread, but on my machine i had to disable zfs-zed due to eating up 100% cpu, 1 core maxed out.
But that might be cause im using old kernel ( couldn't get new one to give me a bootable system yet) + new zfs userland
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.