CephFS snapshots and kernel cephfs continuous I/O
Hello ceph users and devs, I've set up CephFS as a storage which ingests data for multiple hours and snap-schedule hourly. This however causes problems in the case of clients which cannot handle the EAGAIN (Resource busy/temporary unavailable) the kernel gives back during snapshotting. Is there any plans or option to change this behaviour to just wait until the snapshot is done and continue IO as nothing happened? Thanks in advance. -- Alex D. RedXen System & Infrastructure Administration https://redxen.eu/
Hi Caskd, We have a protocol called "Quiesce" that can pause I/O for snapshot consistency [0]. If that is something that might help you, do check it out. [0] https://docs.ceph.com/en/latest/dev/mds_internals/quiesce/ -- *Dhairya Parmar* Associate Software Engineer, CephFS <https://www.redhat.com/>IBM, Inc. On Sun, Jul 7, 2024 at 7:14 PM caskd <caskd@redxen.eu> wrote:
Hello ceph users and devs,
I've set up CephFS as a storage which ingests data for multiple hours and snap-schedule hourly. This however causes problems in the case of clients which cannot handle the EAGAIN (Resource busy/temporary unavailable) the kernel gives back during snapshotting. Is there any plans or option to change this behaviour to just wait until the snapshot is done and continue IO as nothing happened?
Thanks in advance.
-- Alex D. RedXen System & Infrastructure Administration https://redxen.eu/ _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Hi, First, i'd like to thank you for the quick reply.
We have a protocol called "Quiesce" that can pause I/O for snapshot consistency [0]. If that is something that might help you, do check it out.
[0] https://docs.ceph.com/en/latest/dev/mds_internals/quiesce/
This looks exactly like what i'd want. Apparently rbd has "rbd_default_snapshot_quiesce_mode" to control the behaviour for rbd snapshots. Is there no counterpart for MDS? It would be great to control this behaviour directly in the snapshotting mechanism rather than handling it externally as snap_schedule is a mgr module and it would require moving this outside. -- Alex D. RedXen System & Infrastructure Administration https://redxen.eu/
I understand the need to be able to control quiescing from mgr/snap_schule and there's already a feature ticket for this [0] but until that's done, there's no other way I can think of. CCing @Milind Changire <mchangir@redhat.com> @Venky Shankar <vshankar@redhat.com> @Patrick Donnelly <pdonnell@redhat.com>. [0] https://tracker.ceph.com/issues/64507 On Mon, Jul 8, 2024 at 6:33 PM caskd <caskd@redxen.eu> wrote:
Hi,
First, i'd like to thank you for the quick reply.
We have a protocol called "Quiesce" that can pause I/O for snapshot consistency [0]. If that is something that might help you, do check it out.
[0] https://docs.ceph.com/en/latest/dev/mds_internals/quiesce/
This looks exactly like what i'd want. Apparently rbd has "rbd_default_snapshot_quiesce_mode" to control the behaviour for rbd snapshots. Is there no counterpart for MDS? It would be great to control this behaviour directly in the snapshotting mechanism rather than handling it externally as snap_schedule is a mgr module and it would require moving this outside.
-- Alex D. RedXen System & Infrastructure Administration https://redxen.eu/
participants (2)
-
caskd
-
Dhairya Parmar