Merge pull request #13734 from liewegas/wip-jewel-x

qa/suite/upgrade/jewel-x: various fixes

Reviewed-by: Yuri Weinstein <yweinstei@redhat.com>
This commit is contained in:
Sage Weil 2017-03-07 21:25:13 -06:00 committed by GitHub
commit 71db343636
45 changed files with 40 additions and 152 deletions

View File

@ -4,12 +4,9 @@ meta:
upgrade-sequence:
sequential:
- ceph.restart:
daemons: [mon.a, mon.b, mon.c, mds.a, osd.0, osd.1, osd.2, osd.3]
daemons: [mon.a, mon.b, mon.c]
- ceph.restart:
daemons: [mds.a, osd.0, osd.1, osd.2, osd.3]
wait-for-healthy: false
wait-for-osds-up: true
- exec:
mon.a:
- ceph osd set require_kraken_osds
- ceph.restart:
daemons: [osd.0]
- print: "**** done ceph.restart all"

View File

@ -12,6 +12,11 @@ upgrade-sequence:
wait-for-healthy: true
- sleep:
duration: 60
- ceph.restart:
daemons: [mon.b, mon.c]
wait-for-healthy: true
- sleep:
duration: 60
- ceph.restart:
daemons: [osd.0, osd.1]
wait-for-healthy: true
@ -25,18 +30,9 @@ upgrade-sequence:
mon.b:
- sudo ceph osd crush tunables jewel
- print: "**** done ceph osd crush tunables jewel"
- ceph.restart:
daemons: [mon.b, mon.c]
wait-for-healthy: true
- sleep:
duration: 60
- ceph.restart:
daemons: [osd.2, osd.3]
wait-for-healthy: false
wait-for-osds-up: true
- exec:
mon.a:
- ceph osd set require_kraken_osds
- ceph.restart: [osd.3]
- sleep:
duration: 60

View File

@ -1 +0,0 @@
../../../../releases/kraken.yaml

View File

@ -0,0 +1 @@
../../../../releases/luminous.yaml

View File

@ -205,9 +205,4 @@ upgrade-sequence_x:
daemons: [osd.5]
wait-for-healthy: false
wait-for-up-osds: true
- exec:
mon.a:
- ceph osd set require_kraken_osds
- sleep:
duration: 60
- print: "**** done ceph.restart all -x branch mds/osd/mon"

View File

@ -1 +0,0 @@
../stress-split/0-cluster/

View File

@ -1 +0,0 @@
../stress-split/1-jewel-install/

View File

@ -1 +0,0 @@
../stress-split/2-partial-upgrade/

View File

@ -1,17 +0,0 @@
meta:
- desc: |
randomly kill and revive osd
small chance of increasing the number of pgs
overrides:
ceph:
log-whitelist:
- wrongly marked me down
- objects unfound and apparently lost
- log bound mismatch
tasks:
- thrashosds:
timeout: 1200
chance_pgnum_grow: 1
chance_pgpnum_fix: 1
min_in: 4
- print: "**** done thrashosds 3-thrash"

View File

@ -1 +0,0 @@
../stress-split/4-mon/

View File

@ -1 +0,0 @@
../stress-split/6-next-mon/

View File

@ -1 +0,0 @@
../stress-split/8-next-mon/

View File

@ -1,35 +0,0 @@
#
# k=3 implies a stripe_width of 1376*3 = 4128 which is different from
# the default value of 4096 It is also not a multiple of 1024*1024 and
# creates situations where rounding rules during recovery becomes
# necessary.
#
meta:
- desc: |
randomized correctness test for rados operations on an erasure coded pool
using the jerasure plugin with k=3 and m=1
tasks:
- rados:
clients: [client.0]
ops: 4000
objects: 50
ec_pool: true
write_append_excl: false
erasure_code_profile:
name: jerasure31profile
plugin: jerasure
k: 3
m: 1
technique: reed_sol_van
ruleset-failure-domain: osd
op_weights:
read: 100
write: 0
append: 100
delete: 50
snap_create: 50
snap_remove: 50
rollback: 50
copy_from: 50
setattr: 25
rmattr: 25

View File

@ -9,6 +9,9 @@ overrides:
- objects unfound and apparently lost
- log bound mismatch
tasks:
- parallel:
- stress-tasks
stress-tasks:
- thrashosds:
timeout: 1200
chance_pgnum_grow: 1

View File

@ -1 +0,0 @@
../stress-split/4-mon/

View File

@ -1,7 +1,7 @@
meta:
- desc: |
randomized correctness test for rados operations on an erasure coded pool
tasks:
stress-tasks:
- rados:
clients: [client.0]
ops: 4000

View File

@ -0,0 +1 @@
../stress-split/5-finish-upgrade.yaml

View File

@ -1,22 +0,0 @@
meta:
- desc: |
randomized correctness test for rados operations on an erasure coded pool
tasks:
- rados:
clients: [client.0]
ops: 4000
objects: 50
ec_pool: true
write_append_excl: false
op_weights:
read: 100
write: 0
append: 100
delete: 50
snap_create: 50
snap_remove: 50
rollback: 50
copy_from: 50
setattr: 25
rmattr: 25
- print: "**** done rados ec task"

View File

@ -0,0 +1 @@
../stress-split/6-luminous.yaml

View File

@ -1 +0,0 @@
../stress-split/6-next-mon/

View File

@ -1 +0,0 @@
../stress-split/8-next-mon/

View File

@ -1,13 +1,10 @@
meta:
- desc: |
Run ceph on two nodes,
with a separate client-only node.
with a separate client-only node.
Use xfs beneath the osds.
overrides:
ceph:
conf:
mon:
mon warn on legacy crush tunables: false
fs: xfs
roles:
- - mon.a

View File

@ -2,11 +2,11 @@ meta:
- desc: |
install upgrade ceph/-x on one node only
1st half
restart : osd.0,1,2,3,4,5
restart : osd.0,1,2
tasks:
- install.upgrade:
osd.0:
osd.0:
- print: "**** done install.upgrade osd.0"
- ceph.restart:
daemons: [osd.0, osd.1, osd.2, osd.3, osd.4, osd.5]
daemons: [mon.a,mon.b,mon.c,osd.0, osd.1, osd.2]
- print: "**** done ceph.restart 1st half"

View File

@ -9,6 +9,9 @@ overrides:
- objects unfound and apparently lost
- log bound mismatch
tasks:
- parallel:
- stress-tasks
stress-tasks:
- thrashosds:
timeout: 1200
chance_pgnum_grow: 1

View File

@ -1,9 +0,0 @@
meta:
- desc: |
restart mon.a so it is upgraded to -x
tasks:
- ceph.restart:
daemons: [mon.a]
wait-for-healthy: false
wait-for-osds-up: true
- print: "**** done ceph.restart mon.a"

View File

@ -2,7 +2,7 @@ meta:
- desc: |
run randomized correctness test for rados operations
generate write load with rados bench
tasks:
stress-tasks:
- full_sequential:
- radosbench:
clients: [client.0]

View File

@ -1,7 +1,7 @@
meta:
- desc: |
run basic cls tests for rbd
tasks:
stress-tasks:
- workunit:
branch: jewel
clients:

View File

@ -1,7 +1,7 @@
meta:
- desc: |
run basic import/export cli tests for rbd
tasks:
stress-tasks:
- workunit:
branch: jewel
clients:

View File

@ -1,7 +1,7 @@
meta:
- desc: |
librbd C and C++ api tests
tasks:
stress-tasks:
- workunit:
branch: jewel
clients:

View File

@ -1,8 +1,8 @@
meta:
- desc: |
randomized correctness test for rados operations on a replicated pool,
randomized correctness test for rados operations on a replicated pool,
using only reads, writes, and deletes
tasks:
stress-tasks:
- full_sequential:
- rados:
clients: [client.0]

View File

@ -1,7 +1,7 @@
meta:
- desc: |
randomized correctness test for rados operations on a replicated pool with snapshot operations
tasks:
stress-tasks:
- full_sequential:
- rados:
clients: [client.0]

View File

@ -0,0 +1,9 @@
tasks:
- install.upgrade:
osd.3:
client.0:
- ceph.restart:
daemons: [osd.3, osd.4, osd.5]
wait-for-healthy: false
wait-for-osds-up: true

View File

@ -0,0 +1 @@
../../../../releases/luminous.yaml

View File

@ -1,9 +0,0 @@
meta:
- desc: |
restart mon.b so it is upgraded to -x
tasks:
- ceph.restart:
daemons: [mon.b]
wait-for-healthy: false
wait-for-osds-up: true
- print: "**** done ceph.restart mon.b 6-next-mon"

View File

@ -3,7 +3,6 @@ meta:
librbd python api tests
tasks:
- workunit:
branch: jewel
clients:
client.0:
- rbd/test_librbd_python.sh

View File

@ -2,7 +2,7 @@ meta:
- desc: |
swift api tests for rgw
tasks:
- rgw:
- rgw:
client.0:
default_idle_timeout: 300
- print: "**** done rgw 9-workload"

View File

@ -1,12 +0,0 @@
meta:
- desc: |
restart mon.c so it is upgraded to -x
as all mon were upgrated, expected ceph cluster reach quorum
tasks:
- ceph.restart:
daemons: [mon.c]
wait-for-healthy: false
wait-for-osds-up: true
- print: "**** done ceph.restart mon.c 8-next-mon"
- ceph.wait_for_mon_quorum: [a, b, c]
- print: "**** done wait_for_mon_quorum 8-next-mon"