Commit Graph

36 Commits

Author SHA1 Message Date
Yuri Weinstein
96e1774462 changed idle_timeout to 300 2014-04-08 10:11:29 -07:00
Sage Weil
39166b2d34 upgrade/dumpling-x: run dumpling version of s3tests
Signed-off-by: Sage Weil <sage@inktank.com>
2014-04-03 14:22:32 -07:00
Sage Weil
5651ee8131 upgrade/dumpling-x/parallel: restart after cuttlefish->dumpling upgrade
Fixes: #7824
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-27 17:58:09 -07:00
Sage Weil
798daf5886 upgrade/dumpling-x/stress-split: set fastcgi idle timeout to 2 min
30 seconds isn't enough when combined with osd thrashing.

Fixes: #7702
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-20 16:51:03 -07:00
Sage Weil
6fd5ca9cff upgrade/dumpling-x/parallel: test cuttlefish->dumpling starting point
Maybe this will trigger #7519.  And anyway, it's good coverage.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-18 21:34:25 -07:00
Sage Weil
99f4e9dda3 symlink all distros facets to a common set of 'supported' targets
This makes it easier to maintain the supported/tested targets.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-18 13:39:13 -07:00
Sage Weil
7597172695 upgrade/dumpling-x/stress-split: tolerate legacy crush tunables
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-15 09:18:33 -07:00
Yuri Weinstein
752380dc96 added print task 2014-03-14 15:46:04 -07:00
Sage Weil
22c461b501 upgrade/dumpling-x: more mon scrub whitelist
There are two log messages; whitelist them both

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-14 08:31:59 -07:00
Sage Weil
8b4c8cbf1a upgrade/dumpling-x: whitelist mon scrub mismatch
A mixed dumpling/firefly cluster will have each mon independently encode
the full map.  Firefly now puts feature bits in the Incremental so that
we encode the "right" way, but dumpling doesn't have that capability
yet, and we don't plan to backport it.

Fixes: #7704
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-13 16:27:56 -07:00
Sage Weil
efdee0d31c upgrade/dumpling-x: upgrade final client node for final test
This means the final workloads are running firefly code.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-12 15:37:13 -07:00
Sage Weil
7fb5e14691 upgrade/dumpling-x: full librados tests after full upgrade
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-12 14:49:54 -07:00
Sage Weil
cfbbcf7c08 upgrade/dumpling-x/stress-split: use dumpling workunits
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-11 19:16:39 -07:00
Yuri Weinstein
72094a8586 fixed final_load problem 2014-03-11 15:40:58 -07:00
Sage Weil
728bd3c551 upgrade/dumpling-x: upgrade mon before osd
The MDS needs to see the TMAP2OMAP feature bit in the osdmap.  THe mons
have to encode the feature bits passed by the new osds in order for that
to happen.  Ergo, upgrade mons first.

This is only important for the MDS and for users who want to use erasure
coded pools or tiering without restarting the osds *again* after the mons
get upgraded.

Fixes: #7679
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-11 12:11:25 -07:00
Sage Weil
2cf0908946 upgrade/dumpling-x: fix order
Make it match the filename.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-09 14:27:09 -07:00
Sage Weil
9db665644e upgrade/dumpling-x: do not thrash primary-affinity
We can't do it on a mixed cluster

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-07 15:46:14 -08:00
Josh Durgin
9109008ad7 stress-split: use dumpling version of rbd.py tests
The firefly version has a test that depends on new behavior.

Fixes: #7643
Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
2014-03-07 13:03:07 -08:00
Zack Cerza
fbb05b73bc Remove the mds-mon-osd upgrade sequence
It wasn't actually supported.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-03-07 10:20:47 -06:00
Zack Cerza
136775b401 Don't warn on legacy crush tunables
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-03-06 19:51:32 -06:00
Josh Durgin
bda2fd3d89 upgrade: fix typo in subsection name
It's referred to as workload, not workunit, by the parallel task earlier.

Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
2014-03-06 16:04:49 -08:00
Josh Durgin
487be43765 upgrade: fix indentation for final-workload
Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
2014-03-06 16:04:40 -08:00
Sage Weil
022a6a3ebc Revert "The parallel task expects dicts here..."
This reverts commit 79e348320f.
2014-03-06 17:04:29 -08:00
Zack Cerza
79e348320f The parallel task expects dicts here...
We should fix that at some point.

Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
2014-03-06 18:32:00 -06:00
Josh Durgin
7b497e1fe2 get test-upgrade-firefly.sh from the dumpling branch
it doesn't exist elsewhere
2014-03-06 14:27:47 -08:00
Josh Durgin
7d1e97f771 Revert "Update rados_api_tests.yaml"
This needs to reference dumpling now.

This reverts commit 8e88922581.
2014-03-06 14:27:47 -08:00
Sage Weil
51461432ae upgrade/dumpling-x: more indentation
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-06 15:29:48 -08:00
Sage Weil
0bf78e6bac upgrade/dumpling-x: fix indentation
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-06 15:27:50 -08:00
Warren Usui
182667ae44 Fix parallel workunit issues.
Fixed the spacing of workunits inside 2-workload and 4-final-workland
files.  Removed parallel final-workload reference from dumpling.yaml.

Fixes: 7633
Signed-off-by: Warren Usui <warren.usui@inktank.com>
2014-03-06 16:34:32 -06:00
Sage Weil
c3766c582d upgrade/dumpling-x: skip rados api tests that fail against firefly OSDs
Fixes: #7584
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-05 12:38:55 -08:00
Sage Weil
f0ac0f675d upgrade/dumpling-x/parallel: do all final workloads
Run all of the final workloads in succession instead of in separate tests.
Since we are not concerned about timing in this case, only that the
cluster is functioning at the end, this cuts down our test matrix by a
factor of 4.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-05 09:30:34 -08:00
Sage Weil
bce605d392 upgrade/dumpling-x: do not upgrade client host
If we are running against a mixed cluster for the duration of the upgrade,
we need to make sure we are running the *old* client tests against the
cluster.  Do not upgrade client.0.

Fixes: #7574
Signed-off-by: Sage Weil <sage@inktank.com>
2014-03-05 09:27:41 -08:00
wusui
fb510c2942 Merge pull request #12 from ceph/wip-final-workload-wusui
Fix dumpling-x upgrade suite.
2014-03-04 16:09:08 -08:00
Warren Usui
1929196d7e Fix dumpling-x upgrade suite.
final-workload references were added (replacing second task
reference)

Fixes: 7606
Signed-off-by: Warren Usui <warren.usui@inktank.com>
2014-03-04 15:45:49 -08:00
Yuri Weinstein
8e88922581 Update rados_api_tests.yaml
Removed branch ref to fix http://tracker.ceph.com/issues/7584
2014-03-03 16:07:56 -08:00
Sage Weil
474d86ca28 upgrade: dumpling-next -> dumpling-x
Make the final upgrade step unspecified so that it can be specified at
schedule time.  This let us test dumpling -> firefly, dumpling->emperor, or
whatever else we want.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-02-20 09:19:21 -08:00