PANIC: zfs: attempting to increase fill beyond max

[lenny47]

Does anyone know what could cause this host crash:

Oct 13 00:24:23 pve1 kernel: PANIC: zfs: attempting to increase fill beyond max; probable double add in segment [bd41b39000:bd41c9f000]

We have a linux guest VM running multi-day simulations. The guest reported the following:

Oct 13 00:25:09 twistie kernel: sd 1:0:0:1: reservation conflict
Oct 13 00:25:09 twistie kernel: sd 1:0:0:1: [sdb] tag#153 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_OK cmd_age=0s
Oct 13 00:25:09 twistie kernel: sd 1:0:0:1: [sdb] tag#153 CDB: Write(10) 2a 00 27 e6 fc b0 00 00 20 00
Oct 13 00:25:09 twistie kernel: blk_update_request: critical nexus error, dev sdb, sector 669449392 op 0x1WRITE) flags 0x100000 phys_seg 4 prio class 0
Oct 13 00:25:09 twistie kernel: sdb1: writeback error on inode 603980385, offset 5341184, sector 669449400
Oct 13 00:25:14 twistie kernel: sdb1: writeback error on inode 603980385, offset 5345280, sector 669449424
Oct 13 00:25:14 twistie kernel: sd 1:0:0:1: reservation conflict
Oct 13 00:25:14 twistie kernel: sd 1:0:0:1: [sdb] tag#158 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_OK cmd_age=0s
Oct 13 00:25:14 twistie kernel: sd 1:0:0:1: [sdb] tag#158 CDB: Write(10) 2a 00 27 e6 fc c8 00 00 28 00

Heaps of space in zpool:

NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
rpool 7.34T 4.28T 3.06T - - 11% 58% 1.00x ONLINE -

My current guess is that an online resize of sdb disk image earlier in the day looked good but at the point the additional space was used .. crash Info on resize method:

1. Resized disk image in Proxmox gui (via "disk action->resize" menu)
2. parted /dev/sdb (then increase partition 1 to 100%)
3. xfs_growfs -d /dev/sdb1
4. df shows increased capacity

Anyone know if online resize caused this (wrong method / missing step? or online just can't be used?) or if something else?

Thanks in advance for any help.