DMESG Errors

[quartzeye]

I had a power failure that kicked the server off and since I see the following errors in the logs. I have done a fsck on all the containers and in all the VM's. I have also forced a fsck on reboot of the Proxmox node but the errors just keep popping up. All the drives seem to be checking as good and none are out of space. I obviously think that something was corrupted by the outage but I have not found any info that actually helps me figure out what is causing the error and how to fix it.

[ 5357.336689] blk_update_request: I/O error, dev loop3, sector 2387896 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 5478.422743] blk_update_request: I/O error, dev loop5, sector 2633552 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 5478.441856] blk_update_request: I/O error, dev loop5, sector 2633584 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 5957.392034] blk_update_request: I/O error, dev loop3, sector 2387928 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 5957.411762] blk_update_request: I/O error, dev loop3, sector 2387960 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 8357.352069] blk_update_request: I/O error, dev loop5, sector 2633648 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 9558.228513] blk_update_request: I/O error, dev loop6, sector 2434424 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 9558.247853] blk_update_request: I/O error, dev loop6, sector 2434456 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0
[ 9558.267168] blk_update_request: I/O error, dev loop6, sector 2434488 op 0x9WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0

Any guidance here would be appreciated.

 

[t.lamprecht]

Hi,

what CTs are you using? Are some of them using loop devices (maybe indirectly through snaps or so).
This could be from a VM/CT doing a fstrim..
If it'd happen more frequently I'd suggest turning VM/CTs off to narrow down which one it could be, but as is this happens about every hour a few times, so probably not possible to turn some off that long.. Maybe you can try doing some fstrim and/or write/delete operations in VMs/CTs to see if it triggers that.