upgrade/mixed-mons: run cuttlefish tests against (2 cuttlefish, 1 dumpling) mon cluster

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade: fix mixed workloads

We are running cuttlefish/dumpling mix.  Client node still has cuttelfish.
Use cuttelfish workunits.

add multi-region tests

Signed-off-by: Joe Buck <jbbuck@gmail.com>

Add Rhel 6.4 and Centos 6.3 to ceph-deploy test suite.

Per issue #6014 and #6015.

Signed-off-by: Sandon Van Ness <sandon@inktank.com>

Add yaml for rhel 6.4 in distros.

Signed-off-by: Sandon Van Ness <sandon@inktank.com>

rados/monthash: fix 5925 yaml

Signed-off-by: Sage Weil <sage@inktank.com>

added upgrade-dumpling suite to test point releases for dumpling

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

added workunit: branch:dumpling wherever applicable

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

upgrade from cuttlefish to next and dumpling to next

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

upgrade cuttlefish to dumpling, instead of next in upgrade-fs suite

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

renamed the folders and files appropriately

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

replace cuttlefish with dumpling in upgrade suites

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

modified upgrade tasks to run from cuttlefish to dumpling, dumpling to next
and bobtail to dumpling

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

monthrash workloads: Add task to reproduce 5925 with extra logging

Signed-off-by: David Zafman <david.zafman@inktank.com>

upgrade-parallel: test on multiple target distros

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade: do not wait for full quorum on mixed-version mon cluster

It won't happen

upgrade-parallel/rgw: use different client for final run

rgw task doesn't like to be stopped, then started as the same client.
That is probably fixable (it's just the archive/apache.$client dir) but
I didn't look closely.  Avoid the issue by using different client ids
here.

Signed-off-by: Sage Weil <sage@inktank.com>

Debian ceph-deploy test. Added more supported distro yamls.

Added debian 7.0 (wheezy) testing for ceph-deploy. I also added
yaml files for other distro's supported by downburst that we can
test on including older ones which may or may not actually get
used.

Signed-off-by: Sandon Van Ness <sandon@inktank.com>

added rgw task before swift

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

rados/.../cephtool: let us run test_daemon.sh too

This needs the mon to be mon.a; rename it.

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel/stress-split: only upgrade on the first node

In particular, we need the client and half the osds to stay at cuttlefish.

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel: run cuttlefish api tests against cuttlefish/dumpling cluster

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-fs: specify fs

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel: run cuttlefish s3tests against hybrid cluster

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel: restart mon after starting osd thrasher

The tell mon.* currently blocks forever if all mons aren't up.

Merge branch 'master' of github.com:ceph/ceph-qa-suite

Test multi distro on ceph-deploy.

Test multiple distro's (ubuntu, centos, rhel, fedora) on
ceph-deploy

Signed-off-by: Sandon Van Ness <sandon@inktank.com>

rados: add rest-api test

Signed-off-by: Sage Weil <sage@inktank.com>

Merge branch 'master' of github.com:ceph/ceph-qa-suite

added fs, rbd and rgw tasks to upgrade from
bobtail and point releases[v0.61.5, v0.61.6] to cuttlefish.

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

rados/thrash: add in ext4

rados: add thrasher which causes map gaps

Signed-off-by: Samuel Just <sam.just@inktank.com>

Add distro yamls.

Should be no harm to get these in now.

To run on all distros the distros dir should just be symlinked in
the folder that also contains tasks, IE:

suites/ceph-deploy/fs

Or make a distro's folder and symlink individual distro's so you
have:

suites/ceph-deploy/fs/distros/centos.yaml
suites/ceph-deploy/fs/distros/fedora.yaml
suites/ceph-deploy/fs/distros/rhel.yaml
suites/ceph-deploy/fs/distros/ubuntu.yaml

etc...

Signed-off-by: Sandon Van Ness <sandon@inktank.com>

ceph-deploy: remove workunit branch options

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel/rados: fix mon restarts

upgrade-parallel: fix rgw syntax

Signed-off-by: Sage Weil <sage@inktank.com>

removed ceph-deploy branch option to make it run against arbitrary branch

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

removed ceph-deploy branch option
so the nightlies can run ceph-deploy against any arbitrary branch

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

upgrade-cuttlefish: run on xfs, not /

modified workunit branch to cuttlefish

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

upgrade-cuttlefish: test upgrades to new point releases

This only covers rados tests currently.  We should expand this to include rgw,
rbd, and cephfs workloads.

Signed-off-by: Sage Weil <sage@inktank.com>

modified the task yaml to use modified ceph.restart and wait_for_mon_quorum

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

upgrade-parallel/stress-split: fix osd names

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade-parallel: add stress-split collection

upgrade-parallel: do staggered updates

added tasks in the yaml

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

Merge branch 'master' of github.com:ceph/ceph-qa-suite

removed a blank line

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

rados: trim more aggressively

rados: keep fewer osdmaps around

Signed-off-by: Sage Weil <sage@inktank.com>

Merge branch 'master' of github.com:ceph/ceph-qa-suite

task yamls for upgrade-parallel suite

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

Add more stressful thrashing by doing more pg splitting

Signed-off-by: David Zafman <david.zafman@inktank.com>

big: run for 1 hour

rados: add in the mon/caps.sh tests

modified the workload for mixed-mon upgrade suite

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

reorg kernel -> krbd + kcephfs

Signed-off-by: Sage Weil <sage@inktank.com>

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>

big: reenable big cluster

small and medium are passing.

fs: run everything on btrfs, not /

nfs: run osds on btrfs

I hit an ext4 xattr hang; avoid that.

fs: disable ceph-deploy test

Leave this to teh ceph-deploy suite

Signed-off-by: Sage Weil <sage@inktank.com>

big: enable medium cluster

Signed-off-by: Sage Weil <sage@inktank.com>

fs: disable maxfid smbtorture test

This fails occasionally, unsure why.

Signed-off-by: Sage Weil <sage@inktank.com>

added upgrade task for mixed mon test

Signed-off-by: tamil <tamil@tamil-VirtualBox.(none)>

Added task yaml for partial-upgrade of osds

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

restart rgw after upgrade for rgw tests

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

rados/mon_thrash: inject random delays in mon messages

More useful than osd delays for this collection

rados/monthrash: reduce matrix size a bit

we don't need so many osd workloads

rados/monthrash: refactor a bit

Signed-off-by: Sage Weil <sage@inktank.com>

rados: expand mon thrashing tests

- test sync (full and recent)
- test multiple mons down
- test other cluster sizes (3 and 9)

Signed-off-by: Sage Weil <sage@inktank.com>

big: disable all but small until it passes

Revert "rbd: reeanble iozone test, just to see"

This reverts commit 81862ad907e9ba94f0a2dfc9ed365b68f305cd8d.

Yeah, still broken.

kernel: simple_1tb -> simple_big

1TB is too big for plana

test_o_trunc.sh moved

rbd: reeanble iozone test, just to see

See #2689

fs: add o_trunc test on ceph-fuse

Signed-off-by: Sage Weil <sage@inktank.com>

kernel: test o_trunc

Signed-off-by: Sage Weil <sage@inktank.com>

upgrade/rgw: use bobtail s3tests

upgrade/rgw: mds restart order doesn't matter

upgrade/rgw-double: start rgw

rbd: remove kernel rbd test from rbd suite

It's already present in the kernel suite.

kernel: specify fs (and use raw disks) for rbd-nomount tests

Otherwise simple 1tb test fills up the root disks.  :(

upgrade: ignore mds restart order on rbd, rados tests

Not relevant to rbd; put these in the fs upgrade matrix.

rgw: move radosgw-admin test into rgw suite

kernel: add instal, ceph tasks to rbd-nomount collection

big: timebox test to two hours

big: thrash, and whitelist wrongly marked me down

upgrade: move fs upgrade tests to a separate suite

For now.. until we figure out why they keep failing!

kernel: add rbd test to write a 1TB image

Motivated by #5454

marginal: restructure the multimds collection

- two clients types (ceph-fuse and kclient)
- two clients in each test
- separate out tasks

kernel: reorg kernel rbd tests; add a few missing scripts

Move all the simple workunits into a separate collection, and test with
and without msgr failure injection.

There were a few tests in the workunits dir that we weren't running.

rbd/rgw upgrade suites to test upgrade from bobtail to cuttlefish to next

Signed-off-by: tamil <tamil.muthamizhan@inktank.com>

rados: dump-stick: remove configs from yaml; let task set them

The mon osd report timeout was wrong; see #5440.

ceph-deploy: no need for ceph-fuse for non fs workunits

ceph-deploy: install correct branch for rados api tests

Fixes: #5442
Signed-off-by: Sage Weil <sage@inktank.com>

Revert "upgrade: add double-hop bobtail -> cuttlefish -> next fs tests"

This reverts commit fbdb7774bb381ff1426650c47cdb36282c4c77d2.

We don't care about bobtail fs.

upgrade: add rados-double suite (bobtail -> cuttlefish -> next)

upgrade/rados: refactor, expand matrix, start with cuttlefish

upgrade/fs: make single hop cuttlefish -> next (not bobtail -> next)

The bobtail ceph-fuse is buggy and these tests keep failing e.g. on umount,
but we don't really care about fixing the bobtail fs bugs.