Bug #2099
closedmessenger: unexpected socket state (4)
0%
Description
Running tests defined by the YAML file below. Note that branch
wip-messenger is 107a8aaf21d01ee6cbc7a638faf1328f2bd98a5f with
some changes that I believe to be sound (but I could be wrong
about that...).
Anyway, the following WARN trigger:
Feb 21 15:54:20 sepia18 kernel: [13246.854753] libceph: client0 fsid 01ae2c34-ef7b-40ef-980f-f04538cb4415
Feb 21 15:54:20 sepia18 kernel: [13246.856029] libceph: mon0 10.3.14.144:6789 session established
Feb 21 15:55:37 sepia18 kernel: [13323.810028] libceph: tid 810 timed out on osd1, will reset osd
Feb 21 15:55:37 sepia18 kernel: [13323.810324] libceph: tid 874 timed out on osd0, will reset osd
Feb 21 15:55:37 sepia18 kernel: [13323.818190] ------------[ cut here ]------------
Feb 21 15:55:37 sepia18 kernel: [13323.818207] WARNING: at /srv/autobuild-ceph/gitbuilder.git/build/net/ceph/messenger.c:222 ceph_state_change+0x6f/0x160 [libceph]()
Feb 21 15:55:37 sepia18 kernel: [13323.818210] Hardware name: PDSMi
Feb 21 15:55:37 sepia18 kernel: [13323.818212] unexpected socket state (4)
Socket state 4 is TCP_FIN_WAIT1, which, according to tcp_fin() comments:
/* This case occurs when a simultaneous close
* happens, we must ack the received FIN and
* enter the CLOSING state.
*/
So maybe this can be treated as the same as TCP_CLOSE but I'm not sure.
Here is the YAML file. The machines that were involved were sepia17,
sepia18, and sepia19; sepia18 is the client. I'm sorry, I don't know
more precisely right now which test was running.
nuke-on-error:
roles:
- [mon.a, mon.c, osd.0]
- [mon.b, mds.a, osd.1]
- [client.0]
kernel:
osd:
branch: wip-messenger
client:
branch: wip-messenger
tasks:
- ceph:
- kclient:
- workunit:
client.0:
- misc/trivial_sync.sh
- suites/blogbench.sh
- suites/dbench.sh
- suites/iozone.sh
- suites/pjd.sh
- suites/tiobench.sh
- suites/fsstress.sh
- kernel_untar_build.sh
- rbd/copy.sh
- rbd/import_export.sh
- rbd/test_librbd.sh