Should we set `failmode=continue` for ZFS pools?

[kvaps]

By default, ZFS freezes I/O on ZVOLs when the underlying pool has failed. As a result, DRBD devices on top may get stuck: they can keep running and report UpToDate even though I/O is no longer progressing.

To prevent this, set the pool failmode to continue:

zpool set failmode=continue data

We should consider setting this flag automatically in the linstor ps cdp command and update the documentation accordingly.

cc @WanzenBug @ghernadi

[WanzenBug]

FYI, next LINSTOR release will add the option to pass arguments to zpool create (as well as pvcreate, vgcreate, etc...). We will probably not set it by default within LINSTOR, but you can manually set it on linstor ps cdp, or perhaps the LINSTOR Operator will use it by default.

[kvaps]

@WanzenBug thank you! I'd like to get some recomenations on setting on keep this option unset for DRBD

[WanzenBug]

Internal consensus was: for DRBD you always want to set failmode=continue.

[kvaps]

@WanzenBug thank you for this note!