RepoMirrors/ceph
1
0
Fork 0
mirror of https://github.com/ceph/ceph synced 2024-12-28 14:34:13 +00:00
Code Issues Packages Projects Releases Wiki Activity
2,086 Commits 211 Branches 0 Tags 1.4 GiB
f8a2a53c59
Commit Graph

2086 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Dan Mick
381d4aa873 lock.py: request only rsa keys from ssh-keyscan 
New versions of ssh-keyscan return two keytypes if possible; this
breaks the comparison of "number of lines of output from keyscan"
to "number of hosts we request keys from".  Fix by asking for only
one type of key (as older ssh-keyscans did).

Fixes: #7164
Signed-off-by: Dan Mick <dan.mick@inktank.com>
 2014-01-16 12:51:39 -08:00
Samuel Just
66312f7ea1 ceph_manager: in test_map_discontinuity, delay killing osd for 20s 
This should give enough time for the replica to get the message
marking them clean.

Signed-off-by: Samuel Just <sam.just@inktank.com>
 2014-01-16 09:57:44 -08:00
Zack Cerza
53fc2d93dd Log a warning when killing long-running jobs. 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-16 10:53:53 -06:00
Zack Cerza
769ef8a960 Kill jobs that run for over 3 days (configurable) 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-16 10:38:39 -06:00
Sage Weil
69d4962759 Merge pull request #174 from ceph/wip-7171 
thrashosds: change min_in from 2 -> 3
 2014-01-10 11:08:31 -08:00
Sage Weil
9a9ee610bd valgrind: ignore tcmalloc uninitialized memory 
This is the main source of noise when running valgrind +
tcmalloc.  Apparently there are other issues, so I think we
still need the notcmalloc gitbuilder, but this gets us part of
the way.

Signed-off-by: Sage Weil <sage@inktank.com>
 2014-01-10 11:02:16 -08:00
Sage Weil
495f2163a8 thrashosds: change min_in from 2 -> 3 
See #7171. In rare cases CRUSH can't handle it when only 2/6 of
the OSDs are marked in.  Avoid those situations for now.

Signed-off-by: Sage Weil <sage@inktank.com>
 2014-01-10 11:00:55 -08:00
Zack Cerza
0c339fe427 Merge pull request #169 from ceph/wip-utsrelease 
kernel: use utsrelease string for need_to_install() purposes
 2014-01-10 09:22:40 -08:00
Ilya Dryomov
bf481b9134 kernel: use utsrelease string for need_to_install() purposes 
Currently, to see if a node has rebooted into the right kernel,
need_to_install() compares a given 40-char commit hash with a 7-char
commit hash abbreviation it pulls from the output of 'uname -r'.

gitbuilders can now export UTS_RELEASE kernel version string through
.../$SHA1/version file.  Use this string instead of the 40-char commit
hash and compare it with the output of 'uname -r' directly.  This saves
us the parsing exercise and, more importantly, makes it possible to
install clean tagged kernels using 'tag:' element, which wasn't
possible before because version string of such kernels doesn't have
a commit hash in it.

If version file is unavailable, fallback to the existing way of doing
things.

Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
 2014-01-10 12:26:09 +02:00
Alfredo Deza
4c049b37e6 Merge pull request #173 from ceph/wip-ceph-deploy 
fix ceph-dpeloy tests to use 2x replication
 2014-01-09 05:43:04 -08:00
Sage Weil
1b4368b3cf schedule_suite: 2x replication for ceph-deploy 
This avoids having to update all of the tests.

Signed-off-by: Sage Weil <sage@inktank.com>
 2014-01-08 17:35:10 -08:00
Zack Cerza
c4b9dae4f6 Merge pull request #172 from ceph/wip-teuthvmcon-wusui 
Fix a bug where ctx.config['targets'] was looped through again
 2014-01-07 15:09:39 -08:00
Warren Usui
f32adf055a Fix a bug where ctx.config['targets'] was looped through again 
in connect().  The bug caused vm behavior to happen for a
target if any of the machines in the cluster was a vm.  The code
was also changed to set the key to none only if rsa or dsa keys
were used on a vm.

Fixes: 7113
Signed-off-by: Warren Usui <warren.usui@inktank.com>
 2014-01-07 14:22:57 -08:00
Alfredo Deza
0c9129bf43 error after 15 minutes of waiting for gatherkeys 
Signed-off-by: Alfredo Deza <alfredo.deza@inktank.com>
(cherry picked from commit ad18717505c407f2c8bb995af8fd516b61957661)
 2014-01-06 15:25:33 -05:00
Zack Cerza
5c0496633c Enable reporting of entire runs as dead 2014-01-06 11:08:12 -06:00
Zack Cerza
3cffea4917 Re-raise exceptions caught in the watchdog 2014-01-03 15:45:18 -06:00
Zack Cerza
027929e0cb Use response.text if response.json is None 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 15:08:45 -06:00
Zack Cerza
f92174ff31 Strip stdout lines 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 15:01:31 -06:00
Zack Cerza
68b259fd00 Catch and log unhandled exceptions in the watchdog 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 14:56:46 -06:00
Zack Cerza
c6a9de0445 Add 'emperor' to list of branches with reporting 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 14:45:25 -06:00
Zack Cerza
eaa3e1240e Work around a change in pip 1.5 regarding wheels 
The error message was "pip's wheel support requires setuptools >= 0.8
for dist-info support."

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 12:41:11 -06:00
Zack Cerza
d3afebe19c Be safer when calling ./bootstrap 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2014-01-03 11:55:13 -06:00
Sandon Van Ness
131b218465 Use CentOS Gitbuilder sha1 instead of Fedora for non-ubuntu. 
Signed-off-by: Sandon Van Ness <sandon@inktank.com>
 2014-01-02 18:30:08 -08:00
Alfredo Deza
d88ec89a67 break out of the while loop after 15 minutes 
Signed-off-by: Alfredo Deza <alfredo.deza@inktank.com>
(cherry picked from commit bef6eb74dcaa37b70b1eab4d28bfa10abb0049d0)
Signed-off-by: Zack Cerza <zack@cerza.org>
 2014-01-02 14:48:49 -06:00
Zack Cerza
b4f524ebe4 Sleep once outside of the watchdog loop 
Hopefully this will prevent the double-posting of jobs.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-31 14:25:05 -06:00
Alfredo Deza
fc815f75fc Merge pull request #168 from ktdreyer/readme-formatting 
format bullets in README
 2013-12-31 05:53:51 -08:00
Ken Dreyer
3d895a0f25 format bullets in README 2013-12-30 20:37:53 -07:00
Zack Cerza
56c70e1145 Set the content-type in report_job() 2013-12-30 17:30:14 -06:00
Zack Cerza
ba8f99991a Split out ResultsSerializer.job_info() 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-30 17:30:14 -06:00
Zack Cerza
e985ca2aab Port from httplib2 to requests module 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-30 17:30:14 -06:00
Sage Weil
6b87147590 valgrind.supp: ignore libnss3 leaks 
These just started popping up when I updated the notcmalloc gitbuilder, probably
because of an updated libnss version.   Whitelist it!

Signed-off-by: Sage Weil <sage@inktank.com>
 2013-12-30 10:12:37 -06:00
Ilya Dryomov
d781348fd5 rbd: bump the default scratch size for xfstests to 10G 
autobuild-ceph.git commit 53db7a34aba5 had silently changed the default
elevator from cfq to deadline, which made xfstests 167 very unhappy.
It looks like with deadline and noop elevators it requires a ~6G
scratch partition.  Bump the default scratch image size to 10G.

Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
 2013-12-23 19:54:11 +02:00
Sage Weil
c70e086b35 Revert "valgrind.supp: ignore libnss3 leaks" 
This reverts commit 572dc88a7cc295cb06354e6f004f7ad665b101f4.

This didn't occur on next; I think there may be a real leak on the ceph
side.
 2013-12-22 09:50:12 -08:00
Sage Weil
572dc88a7c valgrind.supp: ignore libnss3 leaks 
These just started popping up.  Probably because I gave the
gitbuilders a kick?

Signed-off-by: Sage Weil <sage@inktank.com>
 2013-12-21 22:21:49 -08:00
SandonV
e13e4562ef Merge pull request #166 from ceph/wip-lockspell-wusui 
Fix spelling error in comment.
 2013-12-20 12:48:42 -08:00
Warren Usui
1810781441 Fix spelling error in teuthology/task/locktest.py comment 2013-12-20 12:39:30 -08:00
Zack Cerza
f8ce69cb9b Add ability to mark jobs as 'dead' 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-20 10:01:03 -06:00
Zack Cerza
e312048ade Allow passing multiple job_ids 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-19 16:43:11 -06:00
Zack Cerza
220779c8c4 Implement single-job killing 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-19 16:32:41 -06:00
Zack Cerza
eeeb6267f9 For teuthology-kill, s/suite/run/ 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-19 16:32:41 -06:00
SandonV
228358ff17 Merge pull request #165 from ceph/wip-7042-fix-wusui 
Do not run local handling fix if local parameter is not found.
 2013-12-19 14:27:16 -08:00
Warren Usui
37815b76d3 Do not run local handling fix if local parameter is not found. 
Fixes: 7042
Signed-off-by: Warren Usui <warren.usui@inktank.com>
 2013-12-19 14:20:12 -08:00
Zack Cerza
dbcef31ba0 Merge pull request #156 from ceph/teuthology-doc-hadoop-wusui 
Added docstrings.  Cleaned up code (broke up long lines, removed unused
 2013-12-19 09:27:14 -08:00
Zack Cerza
006c031182 Merge pull request #164 from ceph/wip-rados 
rados: add in more (optional) op types
 2013-12-19 09:24:21 -08:00
Zack Cerza
d70d1ad76f Merge pull request #160 from ceph/wip-fix-5149-wusui 
Added handling of a 'local' option inside install.py which specifies
 2013-12-19 09:23:36 -08:00
Zack Cerza
9a29c3ef71 Log calls to teuthology-report more verbosely 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-19 10:29:30 -06:00
Zack Cerza
b014c71829 Catch every exception here, for now. 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-19 10:29:30 -06:00
Sandon Van Ness
031be56813 Use saucy gitbuilder for arm package checking. 
Some-how missed it checks both sha1 and package version file
and package version was still the quantal gitbuilder which wont
work as the hardware is down.

This was causing scheduling failures.

Signed-off-by: Sandon Van Ness <sandon@inktank.com>
 2013-12-18 12:38:50 -08:00
Sage Weil
5320db57ce rados: add in more (optional) op types 
Signed-off-by: Sage Weil <sage@inktank.com>
 2013-12-18 11:41:58 -08:00
Zack Cerza
a0eb1a8e8c Use shell=True to call teuthology-report 
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
 2013-12-16 14:22:22 -06:00