qa/tasks: start DaemonWatchdog when ceph starts

* Start DaemonWatchdog when ceph starts
* Drop the DaemonWatchdog starting in mds_thrash.py
* Bring the thrashers in mds_thrash.py into the context

Fixes: http://tracker.ceph.com/issues/10369
Signed-off-by: Jos Collin <jcollin@redhat.com>

diff --git a/qa/tasks/ceph.py b/qa/tasks/ceph.py
index 38c1a3ae4a0263ad542d28e550a749ecd24a4c77..cf8931793d8ab042ed70a2c35be23aca2e74a992 100644 (file)
--- a/qa/tasks/ceph.py
+++ b/qa/tasks/ceph.py
@@ -27,6 +27,7 @@ from teuthology import exceptions
 from teuthology.orchestra import run
 import ceph_client as cclient
 from teuthology.orchestra.daemon import DaemonGroup
+from tasks.daemonwatchdog import DaemonWatchdog
 
 CEPH_ROLE_TYPES = ['mon', 'mgr', 'osd', 'mds', 'rgw']
 DATA_PATH = '/var/lib/ceph/{type_}/{cluster}-{id_}'
@@ -416,6 +417,12 @@ def cephfs_setup(ctx, config):
 
     yield
 
+@contextlib.contextmanager
+def watchdog_setup(ctx, config):
+    ctx.ceph[config['cluster']].thrashers = []
+    ctx.ceph[config['cluster']].watchdog = DaemonWatchdog(ctx, config, ctx.ceph[config['cluster']].thrashers)
+    ctx.ceph[config['cluster']].watchdog.start()
+    yield
 
 def get_mons(roles, ips, cluster_name,
              mon_bind_msgr2=False,
@@ -1672,6 +1679,8 @@ def stop(ctx, config):
         cluster, type_, id_ = teuthology.split_role(role)
         ctx.daemons.get_daemon(type_, id_, cluster).stop()
 
+    ctx.ceph[config['cluster']].watchdog.stop()
+    ctx.ceph[config['cluster']].watchdog.join()
     yield
 
 
@@ -1909,6 +1918,7 @@ def task(ctx, config):
         lambda: create_rbd_pool(ctx=ctx, config=config),
         lambda: cephfs_setup(ctx=ctx, config=config),
         lambda: run_daemon(ctx=ctx, config=config, type_='mds'),
+        lambda: watchdog_setup(ctx=ctx, config=config),
     ]
 
     with contextutil.nested(*subtasks):

diff --git a/qa/tasks/mds_thrash.py b/qa/tasks/mds_thrash.py
index 859446ed57cebf2cdb67d47f82e478e274aea870..a4ef8d9850534e55fb24aa99267df6c14666b28b 100644 (file)
--- a/qa/tasks/mds_thrash.py
+++ b/qa/tasks/mds_thrash.py
@@ -14,7 +14,6 @@ from gevent.event import Event
 from teuthology import misc as teuthology
 
 from tasks.cephfs.filesystem import MDSCluster, Filesystem
-from tasks.daemonwatchdog import DaemonWatchdog
 
 log = logging.getLogger(__name__)
 
@@ -409,29 +408,25 @@ def task(ctx, config):
         status = mds_cluster.status()
     log.info('Ready to start thrashing')
 
-    thrashers = []
-
-    watchdog = DaemonWatchdog(ctx, manager, config, thrashers)
-    watchdog.start()
-
     manager.wait_for_clean()
     assert manager.is_clean()
+
+    if 'cluster' not in config:
+        config['cluster'] = 'ceph'
+
     for fs in status.get_filesystems():
         thrasher = MDSThrasher(ctx, manager, config, Filesystem(ctx, fs['id']), fs['mdsmap']['max_mds'])
         thrasher.start()
-        thrashers.append(thrasher)
+        ctx.ceph[config['cluster']].thrashers.append(thrasher)
 
     try:
         log.debug('Yielding')
         yield
     finally:
         log.info('joining mds_thrashers')
-        for thrasher in thrashers:
+        for thrasher in ctx.ceph[config['cluster']].thrashers:
             thrasher.stop()
             if thrasher.e:
                 raise RuntimeError('error during thrashing')
             thrasher.join()
         log.info('done joining')
-
-        watchdog.stop()
-        watchdog.join()