summaryrefslogtreecommitdiffstats
path: root/qa/tasks/peer.py
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 18:45:59 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 18:45:59 +0000
commit19fcec84d8d7d21e796c7624e521b60d28ee21ed (patch)
tree42d26aa27d1e3f7c0b8bd3fd14e7d7082f5008dc /qa/tasks/peer.py
parentInitial commit. (diff)
downloadceph-19fcec84d8d7d21e796c7624e521b60d28ee21ed.tar.xz
ceph-19fcec84d8d7d21e796c7624e521b60d28ee21ed.zip
Adding upstream version 16.2.11+ds.upstream/16.2.11+dsupstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r--qa/tasks/peer.py90
1 files changed, 90 insertions, 0 deletions
diff --git a/qa/tasks/peer.py b/qa/tasks/peer.py
new file mode 100644
index 000000000..6b19096b1
--- /dev/null
+++ b/qa/tasks/peer.py
@@ -0,0 +1,90 @@
+"""
+Peer test (Single test, not much configurable here)
+"""
+import logging
+import json
+import time
+
+from tasks import ceph_manager
+from tasks.util.rados import rados
+from teuthology import misc as teuthology
+
+log = logging.getLogger(__name__)
+
+def task(ctx, config):
+ """
+ Test peering.
+ """
+ if config is None:
+ config = {}
+ assert isinstance(config, dict), \
+ 'peer task only accepts a dict for configuration'
+ first_mon = teuthology.get_first_mon(ctx, config)
+ (mon,) = ctx.cluster.only(first_mon).remotes.keys()
+
+ manager = ceph_manager.CephManager(
+ mon,
+ ctx=ctx,
+ logger=log.getChild('ceph_manager'),
+ )
+
+ while len(manager.get_osd_status()['up']) < 3:
+ time.sleep(10)
+ manager.flush_pg_stats([0, 1, 2])
+ manager.wait_for_clean()
+
+ for i in range(3):
+ manager.set_config(
+ i,
+ osd_recovery_delay_start=120)
+
+ # take on osd down
+ manager.kill_osd(2)
+ manager.mark_down_osd(2)
+
+ # kludge to make sure they get a map
+ rados(ctx, mon, ['-p', 'data', 'get', 'dummy', '-'])
+
+ manager.flush_pg_stats([0, 1])
+ manager.wait_for_recovery()
+
+ # kill another and revive 2, so that some pgs can't peer.
+ manager.kill_osd(1)
+ manager.mark_down_osd(1)
+ manager.revive_osd(2)
+ manager.wait_till_osd_is_up(2)
+
+ manager.flush_pg_stats([0, 2])
+
+ manager.wait_for_active_or_down()
+
+ manager.flush_pg_stats([0, 2])
+
+ # look for down pgs
+ num_down_pgs = 0
+ pgs = manager.get_pg_stats()
+ for pg in pgs:
+ out = manager.raw_cluster_cmd('pg', pg['pgid'], 'query')
+ log.debug("out string %s",out)
+ j = json.loads(out)
+ log.info("pg is %s, query json is %s", pg, j)
+
+ if pg['state'].count('down'):
+ num_down_pgs += 1
+ # verify that it is blocked on osd.1
+ rs = j['recovery_state']
+ assert len(rs) >= 2
+ assert rs[0]['name'] == 'Started/Primary/Peering/Down'
+ assert rs[1]['name'] == 'Started/Primary/Peering'
+ assert rs[1]['blocked']
+ assert rs[1]['down_osds_we_would_probe'] == [1]
+ assert len(rs[1]['peering_blocked_by']) == 1
+ assert rs[1]['peering_blocked_by'][0]['osd'] == 1
+
+ assert num_down_pgs > 0
+
+ # bring it all back
+ manager.revive_osd(1)
+ manager.wait_till_osd_is_up(1)
+ manager.flush_pg_stats([0, 1, 2])
+ manager.wait_for_clean()