.Reducing the number of active MDS daemons on CephFS no longer causes kernel client's I/O to hang
Previously, reducing the number of active Metadata Server (MDS) daemons on a Ceph File System (CephFS) would cause kernel client's I/O to hang. When this happens, kernel clients were unable to connect to MDS ranks greater than or equal to `max_mds`. This issue has been fixed in this release.
Description of problem:
Running automated scripts for testing rsync module,but IOs were hung for long hours. Different IO are used(dd,fio,crefi,touch)
Version-Release number of selected component (if applicable):
ceph version 12.2.4-4.el7cp (bfc2b497ab362f2b3afa7bd1f9d0053f74b60d66) luminous (stable)
Steps to Reproduce:
1.Setup ceph cluster and mount on ceph-fuse and kernel clients on same mount point
2.Try to do IOs on machine and mount point,use rsync module to sync data from local to mount dir and data on mount dir to local.
IOs were hung
IOs should be successful and sync should happen
Logs of clients and 2 active mdss are attached
looks like ceph_mdsc_open_export_target_session(mdsc, target) return error. the function only return -ENOMEM (unlikely in this case) and -EINVAL. It return -EINVAL when "target >= mdsmap->m_max_mds". did you change max_mds from 2 to 1 during the test?
(In reply to Yan, Zheng from comment #16)
> looks like ceph_mdsc_open_export_target_session(mdsc, target) return error.
> the function only return -ENOMEM (unlikely in this case) and -EINVAL. It
> return -EINVAL when "target >= mdsmap->m_max_mds". did you change max_mds
> from 2 to 1 during the test?
Yes,for previous test cleanup
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
*** Bug 1594760 has been marked as a duplicate of this bug. ***