@prakashsurya I am all in doubts. On the one hand, I believe that the change is
correct (or, at least, is a move in the correct direction). On the other hand,
I have got some contradictory feedback.
It would be great to get more performance testing.
Might Delphix be interested in that? :-)
--
@avg-I how do you want to proceed with this change? should we move forward with
the RTI? or close it if introduces a regression?
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/openzfs/openzfs/pull/438#iss
I am getting some reports that this change actually hurts performance rather
than improves it.
Perhaps this is hardware dependent, perhaps not.
So, please do not commit this just yet.
I wonder if anyone else tested this change and can share any results.
--
You are receiving this because you ar
grwilson approved this pull request.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/openzfs/openzfs/pull/438#pullrequestreview-62475131
--
openzfs-developer
Arch
ahrens approved this pull request.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/openzfs/openzfs/pull/438#pullrequestreview-62161415
--
openzfs-developer
Archiv
avg-I commented on this pull request.
> @@ -1417,6 +1417,13 @@ zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q,
> boolean_t cutinline)
spa_t *spa = zio->io_spa;
zio_type_t t = zio->io_type;
int flags = (cutinline ? TQ_FRONT : 0);
+ zio_priority_t p = zio->io_prio
ahrens approved this pull request.
> @@ -1417,6 +1417,13 @@ zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q,
> boolean_t cutinline)
spa_t *spa = zio->io_spa;
zio_type_t t = zio->io_type;
int flags = (cutinline ? TQ_FRONT : 0);
+ zio_priority_t p = zio->io_priorit
Oh, I wondered about that strange failure...
Rebased now.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/openzfs/openzfs/pull/438#issuecomment-320031628
--
openzfs-d
@avg-I can you rebase this onto the latest `master` branch? it's failing the
automated build/test because there's changes to the `jenkins` directory in
`master` that don't exist in your branch/PR (i.e. commits that exist on
`master` that aren't in your PR branch).
--
You are receiving this bec
In addition to dispatching synchronous zio-s via "high" task queues, I also
modified the configuration of those task queues to match the regular
counterparts. I did that to avoid any performance regressions.
But it could be that the new values are an overkill. Especially, for the
synchronous wr
10 matches
Mail list logo
