[sheepdog-users] cache flush: all or nothing

[Valerio Pachera]

2013/6/6 Liu Yuan <namei.unix at gmail.com>:
> I can't reproduce the problem even with ext3.

I cant tell you something more.
Yesterday I had to shutdown my production cluster (0.6.0) and I wanted
to update it.
There was something wrong with the guest named 'backup' because I
couldn't see its vnc.
The guest is a debian wheezy (so kernel > 3.0) and uses xfs.
Flushing cache manually took loooong.

Now I've been upgrading to 0.6.0_23_gd0b2857 (not yet changed fd
limit, I'm going to do it right now).

The guest 'backup' is still unable to shut down correctly (I still
can't see the vnc after sending shutdown signal by monitor).

It may be related to fd limit.
I hope not related to the vdi :-)
I got some 'ata1.01: device reported invalid CHS sector 0' on the
guest's /var/log/messages.
Let see after the change.


Sheep.log after upgade (note last rows)
Jun 09 18:45:43 [main] md_add_disk(161)
/mnt/ST2000DM001-1CH164_W1E2N5G6/obj, nr 1
Jun 09 18:45:43 [main] md_add_disk(161) /mnt/wd_WMAYP0904279, nr 2
Jun 09 18:45:43 [main] send_join_request(1101) IPv4 ip:192.168.6.41 port:7000
Jun 09 18:45:43 [main] for_each_object_in_stale(403)
/mnt/ST2000DM001-1CH164_W1E2N5G6/obj/.stale
Jun 09 18:45:43 [main] for_each_object_in_stale(403) /mnt/wd_WMAYP0904279/.stale
Jun 09 18:45:47 [main] check_host_env(395) WARN: Allowed open files
1024 too small, suggested 1024000
Jun 09 18:45:47 [main] check_host_env(404) Allowed core file size 0,
suggested unlimited
Jun 09 18:45:47 [main] main(774) sheepdog daemon (version 0.6.0) started
Jun 09 18:45:47 [main] update_cluster_info(877) status = 4, epoch =
10, finished: 0
Jun 09 18:46:07 [main] sd_check_join_cb(1061) 192.168.6.42:7000: ret =
0x0, cluster_status = 0x4
Jun 09 18:46:07 [main] update_cluster_info(877) status = 4, epoch =
10, finished: 1
Jun 09 18:46:21 [main] sd_check_join_cb(1061) 192.168.6.44:7000: ret =
0x0, cluster_status = 0x1
Jun 09 18:46:21 [main] update_cluster_info(877) status = 1, epoch =
10, finished: 1
Jun 09 18:51:17 [main] modify_event(151) event info for fd 99 not found
Jun 09 18:51:34 [main] main(781) shutdown
Jun 09 18:52:58 [main] md_add_disk(161)
/mnt/ST2000DM001-1CH164_W1E2N5G6/obj, nr 1
Jun 09 18:52:58 [main] md_add_disk(161) /mnt/wd_WMAYP0904279, nr 2
Jun 09 18:52:58 [main] send_join_request(1101) IPv4 ip:192.168.6.41 port:7000
Jun 09 18:52:58 [main] for_each_object_in_stale(403)
/mnt/ST2000DM001-1CH164_W1E2N5G6/obj/.stale
Jun 09 18:52:58 [main] for_each_object_in_stale(403) /mnt/wd_WMAYP0904279/.stale
Jun 09 18:52:59 [main] check_host_env(395) WARN: Allowed open files
1024 too small, suggested 1024000
Jun 09 18:52:59 [main] check_host_env(404) Allowed core file size 0,
suggested unlimited
Jun 09 18:52:59 [main] main(774) sheepdog daemon (version
0.6.0_23_gd0b2857) started
Jun 09 18:52:59 [main] update_cluster_info(877) status = 4, epoch =
10, finished: 0
Jun 09 18:53:01 [main] sd_check_join_cb(1061) 192.168.6.42:7000: ret =
0x0, cluster_status = 0x4
Jun 09 18:53:01 [main] update_cluster_info(877) status = 4, epoch =
10, finished: 1
Jun 09 18:53:05 [main] sd_check_join_cb(1061) 192.168.6.44:7000: ret =
0x0, cluster_status = 0x1
Jun 09 18:53:05 [main] update_cluster_info(877) status = 1, epoch =
10, finished: 1
Jun 09 19:05:34 [main] modify_event(151) event info for fd 26 not found
Jun 09 21:42:15 [main] modify_event(151) event info for fd 23 not found
Jun 09 22:01:35 [main] modify_event(151) event info for fd 59 not found
Jun 09 23:02:19 [main] modify_event(151) event info for fd 78 not found