Hi all,
We have a system (proxmox ve 6.4, ZFS 2.0.7) which ran for about a year and a half without any problems but now has problems getting a ZFS pool mounted/imported. During boot it takes a very long time at Importing ZFS pools using cache file. And eventually the system kernel panics.
I managed to disable the zfs import services using a busybox shell and now i can get into the system.
zpool status takes a long time to complete and shows the zpool ready to import. But importing it results in the same kernel panic (after a while).
I found this thread: https://github.com/openzfs/zfs/issues/10828 and changed this zfs parameter: zfs_multihost_fail_intervals to 0 because i also had the txg_sync process using a lot of CPU cycles.
With the zfs_multihost_fail_intervals=0 the txg_sync is away but the import now runs and is already running for several hours.
I noticed that out of the 32 cpu cores/threads the system has, only 1 core is in 100% io_wait and the rest is idle.
ZFS is reading the disks (12x12TB RAID10) at about 500 KiB/s, around 40KiB/s per disk.
Anyone knows what is going on? Why is import going so slowly and why does it kernel panic in the first place?
Thanks for any advice.
Kind regards,
Caspar
We have a system (proxmox ve 6.4, ZFS 2.0.7) which ran for about a year and a half without any problems but now has problems getting a ZFS pool mounted/imported. During boot it takes a very long time at Importing ZFS pools using cache file. And eventually the system kernel panics.
I managed to disable the zfs import services using a busybox shell and now i can get into the system.
zpool status takes a long time to complete and shows the zpool ready to import. But importing it results in the same kernel panic (after a while).
I found this thread: https://github.com/openzfs/zfs/issues/10828 and changed this zfs parameter: zfs_multihost_fail_intervals to 0 because i also had the txg_sync process using a lot of CPU cycles.
With the zfs_multihost_fail_intervals=0 the txg_sync is away but the import now runs and is already running for several hours.
I noticed that out of the 32 cpu cores/threads the system has, only 1 core is in 100% io_wait and the rest is idle.
ZFS is reading the disks (12x12TB RAID10) at about 500 KiB/s, around 40KiB/s per disk.
Anyone knows what is going on? Why is import going so slowly and why does it kernel panic in the first place?
Thanks for any advice.
Kind regards,
Caspar