Merge pull request ceph#58635 from badone/wip-tracker-50371-rados_api_test-timeout-failures

qa: Restrict rados api tests to large clusters and increase timeout

Reviewed-by: Laura Flores <[email protected]>

Author: yuriw

qa/clusters/2-node-mgr.yaml

@@ -1,3 +1,4 @@
+clusternodes: 2
 roles:
 - [mgr.x, mon.a, mon.c, mds.a, mds.c, osd.0, client.0]
 - [mgr.y, mgr.z, mon.b, mds.b, osd.1, osd.2, osd.3, client.1]

qa/clusters/extra-client.yaml

@@ -1,3 +1,4 @@
+clusternodes: 2
 roles:
 - [mon.a, mon.c, osd.0, osd.1, osd.2]
 - [mon.b, mgr.x, mds.a, osd.3, osd.4, osd.5]
@@ -11,4 +12,4 @@ overrides:
   ceph:
     conf:
       osd:
-        osd shutdown pgref assert: true
+        osd shutdown pgref assert: true

qa/clusters/fixed-1.yaml

@@ -1,3 +1,4 @@
+clusternodes: 1
 overrides:
   ceph-deploy:
     conf:

qa/clusters/fixed-2.yaml

@@ -1,3 +1,4 @@
+clusternodes: 2
 roles:
 - [mon.a, mon.c, mgr.y, osd.0, osd.1, osd.2, osd.3, client.0, node-exporter.a]
 - [mon.b, mgr.x, osd.4, osd.5, osd.6, osd.7, client.1, prometheus.a, node-exporter.b]

qa/clusters/fixed-3-cephfs.yaml

@@ -1,3 +1,4 @@
+clusternodes: 2
 roles:
 - [mon.a, mds.a, mgr.x, osd.0, osd.1]
 - [mon.b, mds.b, mon.c, mgr.y, osd.2, osd.3]

qa/clusters/fixed-3.yaml

@@ -1,3 +1,4 @@
+clusternodes: 2
 roles:
 - [mon.a, mon.c, mgr.x, osd.0, osd.1, osd.2, osd.3]
 - [mon.b, mgr.y, osd.4, osd.5, osd.6, osd.7]

qa/clusters/fixed-4.yaml

@@ -1,3 +1,4 @@
+clusternodes: 4
 roles: 
 - [mon.a, mgr.y, osd.0, osd.4, osd.8, osd.12]
 - [mon.b, osd.1, osd.5, osd.9, osd.13] 
@@ -11,4 +12,4 @@ overrides:
   ceph:
     conf:
       osd:
-        osd shutdown pgref assert: true
+        osd shutdown pgref assert: true

qa/suites/rados/thrash/workloads/rados_api_tests.yaml

@@ -1,3 +1,7 @@
+teuthology:
+  postmerge:
+    # Don't run this test on tiny clusters
+    - if yaml.clusternodes < 4 then reject() end
 overrides:
   ceph:
     log-ignorelist:
@@ -16,6 +20,7 @@ overrides:
       osd:
         osd class load list: "*"
         osd class default list: "*"
+        osd client watch timeout: 240
 tasks:
 - workunit:
     clients:

qa/suites/rados/verify/tasks/rados_api_tests.yaml

@@ -1,3 +1,7 @@
+teuthology:
+  postmerge:
+    # Don't run this test on tiny clusters
+    - if yaml.clusternodes < 4 then reject() end
 overrides:
   ceph:
     log-ignorelist:
@@ -23,7 +27,7 @@ overrides:
       osd:
         osd class load list: "*"
         osd class default list: "*"
-        osd client watch timeout: 120
+        osd client watch timeout: 240
 tasks:
 - workunit:
     timeout: 6h