Merge pull request #16429 from liewegas/wip-jewel-x

qa/suites/upgrade/jewel-x: misc fixes for new health checks
ceph · Jul 20, 2017 · 27e8d75 · 27e8d75
2 parents ed48ed1 + 7102de8
commit 27e8d75
Show file tree

Hide file tree

Showing 5 changed files with 47 additions and 24 deletions.
diff --git a/qa/suites/upgrade/jewel-x/parallel/1-jewel-install/jewel.yaml b/qa/suites/upgrade/jewel-x/parallel/1-jewel-install/jewel.yaml
@@ -11,6 +11,10 @@ tasks:
 - ceph:
     skip_mgr_daemons: true
     add_osds_to_crush: true
+    log-whitelist:
+      - overall HEALTH_
+      - (FS_
+      - (MDS_
 - print: "**** done ceph"
 - install.upgrade:
     mon.a:

diff --git a/qa/suites/upgrade/jewel-x/point-to-point-x/point-to-point-upgrade.yaml b/qa/suites/upgrade/jewel-x/point-to-point-x/point-to-point-upgrade.yaml
@@ -80,6 +80,9 @@ tasks:
    - workload_x
    - upgrade-sequence_x
 - print: "**** done parallel -x branch"
+- exec:
+    osd.0:
+      - ceph osd set-require-min-compat-client luminous
 # Run librados tests on the -x upgraded cluster
 - install.upgrade:
     client.1:
@@ -221,6 +224,5 @@ upgrade-sequence_x:
    - exec:
        osd.0:
          - ceph osd require-osd-release luminous
-         - ceph osd set-require-min-compat-client luminous
    - ceph.healthy:
    - print: "**** done ceph.restart all -x branch mds/osd/mon"
diff --git a/qa/suites/upgrade/jewel-x/stress-split-erasure-code/thrashosds-health.yaml b/qa/suites/upgrade/jewel-x/stress-split-erasure-code/thrashosds-health.yaml
@@ -0,0 +1 @@
+../../../../tasks/thrashosds-health.yaml
diff --git a/qa/suites/upgrade/jewel-x/stress-split/thrashosds-health.yaml b/qa/suites/upgrade/jewel-x/stress-split/thrashosds-health.yaml
@@ -0,0 +1 @@
+../../../../tasks/thrashosds-health.yaml
diff --git a/qa/tasks/ceph.py b/qa/tasks/ceph.py
@@ -1110,13 +1110,17 @@ def run_daemon(ctx, config, type_):
     if config.get('coverage') or config.get('valgrind') is not None:
         daemon_signal = 'term'
 
+    # create osds in order.  (this only matters for pre-luminous, which might
+    # be hammer, which doesn't take an id_ argument to legacy 'osd create').
+    osd_uuids  = {}
     for remote, roles_for_host in daemons.remotes.iteritems():
         is_type_ = teuthology.is_type(type_, cluster_name)
         for role in roles_for_host:
             if not is_type_(role):
                 continue
             _, _, id_ = teuthology.split_role(role)
 
+
             if type_ == 'osd':
                 datadir='/var/lib/ceph/osd/{cluster}-{id}'.format(
                     cluster=cluster_name, id=id_)
@@ -1125,29 +1129,40 @@ def run_daemon(ctx, config, type_):
                     path=datadir + '/fsid',
                     sudo=True,
                 ).strip()
-                try:
-                    remote.run(
-                        args=[
-                            'sudo', 'ceph', '--cluster', cluster_name,
-                            'osd', 'new', osd_uuid, id_,
-                        ]
-                    )
-                except:
-                    # fallback to pre-luminous (hammer or jewel)
-                    remote.run(
-                        args=[
-                            'sudo', 'ceph', '--cluster', cluster_name,
-                            'osd', 'create', osd_uuid,
-                        ]
-                    )
-                if config.get('add_osds_to_crush'):
-                    remote.run(
-                        args=[
-                            'sudo', 'ceph', '--cluster', cluster_name,
-                            'osd', 'crush', 'create-or-move', 'osd.' + id_,
-                            '1.0', 'host=localhost', 'root=default',
-                        ]
-                    )
+                osd_uuids[id_] = osd_uuid
+    for osd_id in range(len(osd_uuids)):
+        id_ = str(osd_id)
+        osd_uuid = osd_uuids.get(id_)
+        try:
+            remote.run(
+                args=[
+                'sudo', 'ceph', '--cluster', cluster_name,
+                    'osd', 'new', osd_uuid, id_,
+                ]
+            )
+        except:
+            # fallback to pre-luminous (hammer or jewel)
+            remote.run(
+                args=[
+                'sudo', 'ceph', '--cluster', cluster_name,
+                    'osd', 'create', osd_uuid,
+                ]
+            )
+            if config.get('add_osds_to_crush'):
+                remote.run(
+                args=[
+                    'sudo', 'ceph', '--cluster', cluster_name,
+                    'osd', 'crush', 'create-or-move', 'osd.' + id_,
+                    '1.0', 'host=localhost', 'root=default',
+                ]
+            )
+
+    for remote, roles_for_host in daemons.remotes.iteritems():
+        is_type_ = teuthology.is_type(type_, cluster_name)
+        for role in roles_for_host:
+            if not is_type_(role):
+                continue
+            _, _, id_ = teuthology.split_role(role)
 
             run_cmd = [
                 'sudo',