From: Loic Dachary <ldachary@redhat.com>
Date: Wed, 16 Nov 2016 09:18:02 +0000 (+0100)
Subject: tests: osd-crush.sh must retry crush dump
X-Git-Tag: v11.1.0~270^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=refs%2Fpull%2F12016%2Fhead;p=ceph.git

tests: osd-crush.sh must retry crush dump

After ceph osd setcrushmap, the script must retry because ceph osd crush
dump races against the update of the crushmap.

Fixes: http://tracker.ceph.com/issues/17919

Signed-off-by: Loic Dachary <loic@dachary.org>
---

diff --git a/src/test/mon/osd-crush.sh b/src/test/mon/osd-crush.sh
index 4fecd4a4d516..982ec904b25f 100755
--- a/src/test/mon/osd-crush.sh
+++ b/src/test/mon/osd-crush.sh
@@ -271,8 +271,19 @@ function TEST_crush_repair_faulty_crushmap() {
     ceph osd setcrushmap -i $empty_map.map || return 1
 
     # should be an empty crush map without any buckets
-    ! test $(ceph osd crush dump --format=xml | \
-           $XMLSTARLET sel -t -m "//buckets/bucket" -v .) || return 1
+    success=false
+    for delay in 1 2 4 8 16 32 64 128 256 ; do
+        if ! test $(ceph osd crush dump --format=xml | \
+                           $XMLSTARLET sel -t -m "//buckets/bucket" -v .) ; then
+            success=true
+            break
+        fi
+        sleep $delay
+    done
+    if ! $success ; then
+        ceph osd crush dump --format=xml
+        return 1
+    fi
     # bring them down, the "ceph" commands will try to hunt for other monitor in
     # vain, after mon.a is offline
     kill_daemons $dir || return 1