Hi
I need to get some clarification on TRIM handling in KVM virtualmachines on TrueNAS and how the commands are intercepted and passed to ZFS.
In this forum thread a user (hege) points out that
OK, I figured this one out. Based on this post, the qemu driver needs the
discard
option set.
and submits it as a bug at JIRA
https://ixsystems.atlassian.net/browse/NAS-122018
This bug was closed with the resolution of Suggestion Not Accepted
A user (@essinghigh) asked if this bug had been looked at and @HoneyBadger responded with:
It’s been merged for the Dragonfish beta on Log in with Atlassian account - let me see if I can prod for a backport to Cobia.
https://ixsystems.atlassian.net/browse/NAS-125642
My Questions
Because the original JRIA issue is closed I cannot ask there and I do not have permission to look at the Dragonfish Beta issue I would like to get the following question answered to clear up how TRIM is implemented in KVM.
- Does unmap/trim get passed and utilised for both thick and thin ZVols?
- Will this change be applied to existing VM configs and will they need a reboot?
- Did this get backported to Cobia?
- did the
detect-zeroes=detect-zeroes
feature get added?, I think this is useful for when ZVols use compression. see the QEMU Man page
Thanks for any information.