Commit Graph

2443 Commits

Author SHA1 Message Date
Yan, Zheng
049c6a99d9 task/nfs: stop nfs server after test finishes more thoroughly
In addition to "exportfs -au", we take a sledgehammer to it
by shutting down all the worker threads and flushing out
the client export table.

Signed-off-by: Yan, Zheng <zyan@redhat.com>
Signed-off-by: Greg Farnum <Greg@inktank.com>
2014-10-20 10:55:25 -07:00
Zack Cerza
51ba2ff551 Better error handling in list_locks()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-20 09:58:25 -06:00
Zack Cerza
9796986236 Fix linter errors
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-20 08:33:42 -06:00
Dan Mick
a8e00f2096 task.ssh_keys(): manage authorized_keys, not keys2
See the issue for a description, but authorized_keys2 is dead
on RHEL7 and dying elsewhere.

Fixes: #9812
Signed-off-by: Dan Mick <dan.mick@inktank.com>
2014-10-17 23:27:30 -07:00
Dan Mick
d3ed4f0756 lock.lock_many(): use passed os_{type,version}, not ctx
Affects teuthology --lock with os_{type,version} in the yaml.

Fixes: #9811
Signed-off-by: Dan Mick <dan.mick@inktank.com>
2014-10-17 22:06:51 -07:00
Zack Cerza
80e0a969ee Instead of setting status='pass', set success=True
Keeping this old behavior will mean that any tasks that are not updated
to use set_status() and still simply set success=False will still be
able to properly report their failures.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-16 13:57:29 -06:00
Zack Cerza
04f8fd85bb Use new functions provided by job_status
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-16 13:51:44 -06:00
Zack Cerza
a75a8ea57f Add new job_status module
Provides methods for getting and setting job status. This is necessary
because we used to just use a boolean called 'success'. In order to stop
reporting infrastructure errors as failures, we need more than just a
boolean.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-16 13:51:44 -06:00
Zack Cerza
db9c92965d Remove setting of unused 's3_branch' value
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-15 12:57:04 -06:00
Zack Cerza
6a5ba7fe6c Stop obsuring stderr when installing with yum
Over time, the code that actually used the buffer we stored stderr in
was removed, leaving us holding error messages and never displaying
them.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-14 09:47:56 -06:00
Zack Cerza
58c22bbb00 Properly quote args that are strings
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-13 16:39:13 -06:00
Zack Cerza
83fce375f1 On SSH connection loss, mark jobs as 'dead'
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-13 16:39:13 -06:00
Zack Cerza
78166c01a4 Use os.path.expanduser() instead of /home
... for locating downburst executable

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-06 14:43:09 -06:00
Zack Cerza
734580d546 Minor fix for vps-only locks
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-03 15:13:02 -06:00
Zack Cerza
2810096cc9 Use arch in lock_machines()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-01 17:15:10 -06:00
Zack Cerza
621a766371 Query arch in lock_many()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-01 17:14:59 -06:00
Zack Cerza
addb3ef73d Use uname -m instead of uname -p
Debian hosts were reporting as 'unknown'

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-01 16:55:55 -06:00
Zack Cerza
cd2860e16c All BaseExceptions should hit failure codepath
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-10-01 14:58:07 -06:00
Zack Cerza
6608323766 Move VersionNotFoundError to exceptions module
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 13:36:36 -06:00
Zack Cerza
05a7944cf2 Move ScheduleFailError to exceptions module
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 13:35:13 -06:00
Zack Cerza
0ef1e12960 Use os_type and os_version in lock_machines()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 10:41:29 -06:00
Zack Cerza
86e934ef25 Query os_type and os_version in lock_many()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 10:41:29 -06:00
Zack Cerza
a962779b15 Add OS.__repr__() and OS.to_dict()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 10:41:29 -06:00
Zack Cerza
a5917b4be9 Move OS class to its own files
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-30 10:41:29 -06:00
Greg Farnum
56223ce98b tasks/nfs: sync before unmounting or unexporting nfs clients/servers
We have some hope that this will resolve #8576 in the Ceph tracker.

Signed-off-by: Greg Farnum <greg@inktank.com>
Reviewed-by: Sage Weil <sage@redhat.com>
2014-09-29 13:15:01 -07:00
Zack Cerza
7781d7f845 Allow mixing of bare-metal machine_types
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-26 12:59:01 -06:00
Zack Cerza
52bf1cf167 Merge pull request #335 from ceph/wip-kernel-downgrade
Fix RPM kernel downgrade.
2014-09-26 11:12:59 -06:00
Greg Farnum
a07f79a79d nfs: more lsof output, on clients too
Signed-off-by: Greg Farnum <greg@inktank.com>
2014-09-25 14:44:20 -07:00
Zack Cerza
a5eed69d1a Revert "Move default os_type to misc.get_distro()"
This reverts commit 3675a2e25c.
2014-09-25 15:22:06 -06:00
Zack Cerza
febc087f44 Revert "Add os_type and os_version args to lock_many()"
This reverts commit f39b6958c4.
2014-09-25 15:22:00 -06:00
Dan Mick
5c83109fbb lock.py: Log failing os-version if it doesn't validate
Fixes: #9597
Signed-off-by: Dan Mick <dan.mick@inktank.com>
2014-09-25 13:56:45 -07:00
Zack Cerza
9e20f1454d Add 'lsb_release -a' parsing back to remote.OS
It can now parse either that or /etc/os-release, and remote.Remote will
attempt the latter first.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-25 09:46:07 -06:00
Zack Cerza
f39b6958c4 Add os_type and os_version args to lock_many()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-25 09:46:07 -06:00
Zack Cerza
3675a2e25c Move default os_type to misc.get_distro()
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-25 09:46:07 -06:00
Zack Cerza
11a67f691e Merge pull request #336 from ceph/wip-nfs
enable lsof when unmounting nfs clients
2014-09-24 15:24:33 -06:00
Greg Farnum
47cdd70b5e enable lsof when unmounting nfs clients
Because the lsof interface is strange, use the verbose flag so we get
some output even if there is no use, and ignore errors because it will
return 1 if there's no filesystem use.

Signed-off-by: Greg Farnum <greg@inktank.com>
2014-09-24 14:13:05 -07:00
Sandon Van Ness
25d4a22b61 Fix RPM kernel downgrade.
Force even if an older version or existing.

Signed-off-by: Sandon Van Ness <sandon@inktank.com>
2014-09-19 14:58:25 -07:00
Zack Cerza
a6ce70d052 Merge pull request #334 from ceph/wip-9519-wusui
Handle list-json downburst failure.
2014-09-19 11:19:27 -06:00
Zack Cerza
dc83719203 Refuse to schedule jobs with 'multi' machine_type
That's not a real machine_type - it's a beanstalkd tube that is used
when a machine_type like 'plana,mira,burnupi' is used.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-19 10:12:59 -06:00
Zack Cerza
98f2e5c8d6 Raise ConnectionLostError when the socket dies
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-19 09:42:36 -06:00
Warren Usui
e37afbb65b Handle list-json downburst failure.
Use check_output instead of Popen and communicate()
Fixes: #9519
Signed-off-by: Warren Usui <warren.usui@inktank.com>
2014-09-18 16:52:31 -07:00
Zack Cerza
3b241620d3 Improve os-release parsing
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-18 13:43:22 -06:00
Zack Cerza
d6515e2c12 Rewrite teuthology-updatekeys
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-18 12:45:51 -06:00
Zack Cerza
b3bb87cd18 Skip hostname test
Due to the workaround put in place recently

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-17 15:48:56 -06:00
Zack Cerza
12ef43b8d0 Re-implement Distribution as OS
It now uses /etc/os-release instead of lsb_release, like the cool kids
are doing.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-17 15:48:56 -06:00
Zack Cerza
deb5dca8db Merge pull request #328 from ceph/wip-8700-wusui
Use docopts to parse daemon-helper
2014-09-17 10:46:23 -06:00
Zack Cerza
c736a52123 Merge pull request #333 from ceph/wip-9495-wusui
Run chef before operations that need it.
2014-09-17 10:41:56 -06:00
Zack Cerza
d277641333 Merge pull request #331 from ceph/wip-transport-none
orchestra: fix ConnectionLostError path in wait()
2014-09-17 10:31:04 -06:00
Zack Cerza
8d458e1b2f Add a workaround for Remote.hostname
Since 'hostname --fqdn' isn't working reliably

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-09-17 10:23:08 -06:00
Warren Usui
329f450ec2 Run chef before operations that need it.
Chef, if it is needed, should be run before kernel tasks
and before all other user defined tasks.  It should not be
run by default.

Fixes: 9495
Signed-off-by: Warren Usui <warren.usui@inktank.com>
2014-09-16 20:23:42 -07:00