I experience a problem with LXC containers. When I shutdown one of them and attempt to start it again, it stucks and node statistics freeze (question marks in GUI, hanging lxc-info processes in htop). Debug start log file looks like this:
PVE versions are these:
And container config...
Nothing upgraded with dist-upgrade:
root@prox1:~# apt dist-upgrade
Reading package lists... Done
Building dependency tree
Reading state information... Done
Calculating upgrade... Done
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
I am experiencing a problem with newest kernel available in pve-no-subscription repository, that's 4.15.18-10-pve. I've upgraded my system with apt update and apt upgrade, current versions are:
root@prox1:~# pveversion -v
proxmox-ve: 5.3-1 (running kernel: 4.15.18-10-pve)
please, do you have any other ideas? That Proxmox instance we were talking about behaves very unstable (random freezes etc.), I would be thankful for any ideas that might solve that problem by successfully moving broken data away.
Thanks a lot
set arc_min to 4GB
set arc_max to 8GB
set zfs_compressed_arc_enabled to 0
And the system freezed completely (IO delay 90%). This was logged into dmesg:
[ 7629.092954] Buffer I/O error on dev zd64, logical block 28727886, async page read
[ 7629.095030] Buffer I/O error...
I have some weird ARC cache problems longer time. It swaps a bit and arcstat says it usually uses max. 1GB of RAM (shortly after boot a lot is used, but then it drops in one second to around 1GB) out of 32GB.
My /sys/module/zfs/parameters/zfs_arc_min and zfs_arc_max equals zero. Swap is on ZFS.
Found out this: "The 'Invalid exchange' error you're seeing is EBADE which was what ZFS uses internally to report a checksum error." (here) - what doesn't make sense, because cat /sys/module/zfs/parameters/zfs_send_corrupt_data still gives 1. It again looks to me that code seen here doesn't...