Commit Graph

52 Commits

Author SHA1 Message Date
Sage Weil
95e6578518 rados/thrash: enable client debug for api tests
These will help catch watch/notify issues...

Signed-off-by: Sage Weil <sage@redhat.com>
2014-12-18 06:41:18 -08:00
Sage Weil
219b72b6fd ms die on skipped message = false for all krbd, kclient tasks
Signed-off-by: Sage Weil <sage@redhat.com>
2014-11-13 21:12:48 -08:00
Loic Dachary
9d2cc2fd2c erasure-code: add thrash-erasure-code to suites/rados
Because it needs the thrasher to have min_in: 4 otherewise k=3,m=1 will
block.

http://tracker.ceph.com/issues/9648 Fixes: #9648

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-10-03 11:59:52 +02:00
Loic Dachary
55e5160d96 erasure-code: fix bugous ec-rados-default.yaml in some suites
The ec-rados-default.yaml was linked to suites as if it was a task
although it is a workload intended for parallel upgrade tests.

The ec-rados-plugin=jerasure-k=2-m=1.yaml task is defined and used
instead, where relevant.

http://tracker.ceph.com/issues/9549 Fixes: #9549

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-09-21 18:58:43 +02:00
Loic Dachary
a624d44566 erasure-code: do not schedule isa workloads just yet
they should be selectively activated, depending the isa plugin availability

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-08-31 10:47:27 +02:00
Loic Dachary
378113cc89 erasure-code: do not schedule lrc workloads just yet
testing after renaming is in progress

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-08-30 18:31:50 +02:00
Loic Dachary
5d2a33d6da erasure-code: add various erasure-code workloads
The erasure-code directory is populated with files describing rados
workloads suitable to test the default erasure code plugins. They are
linked to the rados suite. The pre-existing rados workloads from the
rados suite are replaced with those. Two workloads only differed by the
fact that one of them had less operations than the other. The more
complete set of operations is kept and used for all workloads to
maximize coverage.

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-08-30 11:17:56 +02:00
Loic Dachary
31df59f229 erasure-code: add jerasure k=3 m=1 workload
It validates that http://tracker.ceph.com/issues/9209 is fixed.

Signed-off-by: Loic Dachary <loic-201408@dachary.org>
2014-08-25 19:07:39 +02:00
tamil
367d4da083 added debug messages for radosbench.yaml
Signed-off-by: tamil <tamil.muthamizhan@inktank.com>
2014-07-22 16:15:19 -07:00
Sage Weil
5e880228fd rados/thrash: set osd_max_backfill = 1 sometimes
Hopefully this will help catch leaks in the recovery reservations.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-07-01 16:09:53 -07:00
Samuel Just
77670aeadf thrashers/mapgap.yaml: ignore osd_map_cache_size warning
Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-06-24 10:19:48 -07:00
Samuel Just
09d02abe45 Revert "rados: Fix cache_flush, cache_try_flush, cache_evict and add hit_set_list"
cache_evict is kinda broken at the moment due to snapdir nonsense.  I'm reverting this
test for now until it's fixed so it doesn't obscure the test results.

Bug: 8629
This reverts commit 3dae48c7cd.
2014-06-19 14:25:30 -07:00
David Zafman
3dae48c7cd rados: Fix cache_flush, cache_try_flush, cache_evict and add hit_set_list
Signed-off-by: David Zafman <david.zafman@inktank.com>
2014-06-05 18:24:59 -07:00
David Zafman
99e67abc94 rados/thrash: Add pool_snaps variants
Fixes: #8284

Signed-off-by: David Zafman <david.zafman@inktank.com>
2014-05-05 11:18:14 -07:00
Samuel Just
d2d7b941dd cache-snaps.yaml: set target_max_objects to test snap flush/evict
Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-05-02 17:04:09 -07:00
David Zafman
18334ea3d3 rados/thrash: Fix workload of cache-agent-big
Create a log of objects and operate on some of them
(Initial object creation counts against total operations specified)
Add limit to 20 minutes

Fixes: 8263

Signed-off-by: David Zafman <david.zafman@inktank.com>
2014-05-01 14:39:18 -07:00
David Zafman
b080355902 Fix syntax of erasure coded pool creation
Signed-off-by: David Zafman <david.zafman@inktank.com>
2014-05-01 10:52:57 -07:00
Sage Weil
c623b3dd97 rados/thrash: whitelist 'must scrub before tier agent can activate'
Signed-off-by: Sage Weil <sage@inktank.com>
2014-04-18 06:59:27 -07:00
Sage Weil
b97c3803c5 rados/thrash: rejection backfill reservations sometimes while thrashing 2014-04-03 18:11:01 -07:00
Samuel Just
68343ee193 rados/thrash/workloads: make cache-agent-big use an ec base pool
Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-03-27 11:51:44 -07:00
Samuel Just
c6a22b33e3 rados/thrash/workloads: enable copy_from for the ec workloads
Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-02-23 14:56:10 -08:00
Samuel Just
b9e1341f10 rados/thrash: add ec-radosbench workload
Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-02-18 13:49:12 -08:00
Samuel Just
386650d983 rados/thrash: add ec workloads
Adds ec workloads for snaps, small objects, read-write.

copy-from is not yet enabled.

Signed-off-by: Samuel Just <sam.just@inktank.com>
2014-02-18 13:45:39 -08:00
Sage Weil
5f8eedf01d rados/thrash: test rados against cache pool with agents
Run rados model against a pool with target_max_objects set so that the
tiering agent is working against us.

Signed-off-by: Sage Weil <sage@inktank.com>
2014-02-17 16:41:25 -08:00
Sage Weil
47dcdb324e rados/thrash: add cache + snaps workload
Signed-off-by: Sage Weil <sage@inktank.com>
2014-01-23 07:08:10 -08:00
Sage Weil
a47d16ec8c rados/trash: add cache workload
Signed-off-by: Sage Weil <sage@inktank.com>
2013-12-22 09:46:03 -08:00
Sage Weil
018164ad04 rados/thrash: add small objects ceph_test_rados workload
With small objects we can generate many more ops more quickly and stress
the cluster a bit differently.

Signed-off-by: Sage Weil <sage@inktank.com>
2013-12-19 21:47:19 -08:00
David Zafman
20b7998222 Increase timeout by 50% in thrashers/mapgap.yaml
Seen in /a/teuthology-2013-12-17_23:00:03-rados-next-distro-basic-plana/7188

Signed-off-by: David Zafman <david.zafman@inktank.com>
2013-12-18 16:59:05 -08:00
Sage Weil
39c4674b27 test copy_from when running ceph_test_rados
Skip most of the upgrade suites, except the test that ends up on next.

Signed-off-by: Sage Weil <sage@inktank.com>
2013-10-15 12:20:39 -07:00
Sage Weil
3615c149c2 make xfs.yaml a symlink 2013-10-02 13:30:13 -07:00
Sage Weil
886c4fdd45 make ext4.yaml a symlink 2013-10-02 13:30:13 -07:00
Sage Weil
92cc980be4 make btrfs.yaml a symlink 2013-10-02 13:30:13 -07:00
Sage Weil
30c56cc1cb mark all existing collections with %
This is ignored by old teuthology, but used by newer code to tell us when
to generate a product of all directory items.

Signed-off-by: Sage Weil <sage@inktank.com>
2013-08-28 11:24:29 -07:00
Sage Weil
5849853c5f rados/thrash: add in ext4 2013-07-26 16:27:11 -07:00
Samuel Just
99ae947109 rados: add thrasher which causes map gaps
Signed-off-by: Samuel Just <sam.just@inktank.com>
2013-07-26 10:48:55 -07:00
David Zafman
748d0c80b1 Add more stressful thrashing by doing more pg splitting
Signed-off-by: David Zafman <david.zafman@inktank.com>
2013-07-22 10:20:25 -07:00
Sage Weil
e8c58d3ee9 rados: use 'ms inject intenral delays' during thrashing
This tries to help hit unusual races in the messenger.

Signed-off-by: Sage Weil <sage@inktank.com>
2013-07-16 21:03:17 -07:00
Sage Weil
35f0478e4a use correct branch for admin_socket tests 2013-06-19 10:37:43 -07:00
Sage Weil
d64f264c89 rados: don't use ceph-fuse for workunits
They can run on the local mount/dir.
2013-06-17 16:41:30 -07:00
Samuel Just
526863ee25 remove ext4 from rados thrashing for now
Signed-off-by: Samuel Just <sam.just@inktank.com>
2013-04-23 12:50:17 -07:00
Samuel Just
2230500eda rados/thrashers: enable split testing
Signed-off-by: Samuel Just <sam.just@inktank.com>
2013-03-15 12:23:37 -07:00
Sage Weil
8dcbf8b685 specify deterministic admin socket path for admin_socket.py tests
Otherwise we don't know the path because the default now includes $pid.
2013-02-24 07:51:11 -08:00
Sage Weil
acab068269 add explicit install task before ceph task 2013-02-19 21:02:26 -08:00
Sage Weil
f05f90b527 rados: use 2 instead of 3 machines
Put the client on the same node as some of the osds.  This shouldn't cause
any problems.  It will make timing a bit different, but if anything that
is probably a good thing.

Signed-off-by: Sage Weil <sage@inktank.com>
2013-02-18 13:24:18 -08:00
Sage Weil
8d5ca94331 rados: separate out mon thrasher from thrasher collection
Run a different set of tests.
2013-02-11 07:13:41 -08:00
Samuel Just
d50086020f btrfs.yaml: increase osd op thread timeout
Signed-off-by: Samuel Just <sam.just@inktank.com>
2013-01-28 19:01:27 -08:00
Samuel Just
2f192eaff6 TestRados expects rollback, not snap_rollback
Signed-off-by: Samuel Just <sam.just@inktank.com>
2013-01-23 19:42:45 -08:00
Sage Weil
b7cb1b113f rados/thrash: 3 monitors, so that we can thrash them 2013-01-21 15:09:00 -08:00
Joao Eduardo Luis
f41b5421f8 add mon_thrash task to kernel and rados thrashers collections
Signed-off-by: Joao Eduardo Luis <jecluis@gmail.com>
2013-01-17 21:00:21 -08:00
Sage Weil
cc7bf1bd18 rados: add osd reply delay injection 2013-01-17 20:54:37 -08:00