vstart.sh: fix mgr vs restful command startup race

If the mgr hasn't activated yet we won't have registered the python
commands.  Use 'ceph tell mgr ...' to ensure we block.  (This is only a
problem right after mkfs when the first mgr hasn't started yet.)

Signed-off-by: Sage Weil <sage@redhat.com>

diff --git a/src/vstart.sh b/src/vstart.sh
index 8fb64f9910e643e7f8337a5583afa144ea1e8773..88c35d5e96fd6f3250462261c4bd88aaa29d6025 100755 (executable)
--- a/src/vstart.sh
+++ b/src/vstart.sh
@@ -671,7 +671,9 @@ EOF
         run 'mgr' $CEPH_BIN/ceph-mgr -i $name $ARGS
     done
 
-    if ceph_adm restful create-self-signed-cert; then
+    # use tell mgr here because the first mgr might not have activated yet
+    # to register the python module commands.
+    if ceph_adm tell mgr restful create-self-signed-cert; then
         SF=`mktemp`
         ceph_adm restful create-key admin -o $SF
         RESTFUL_SECRET=`cat $SF`