From: Rishabh Dave <ridave@redhat.com>
Date: Wed, 13 Apr 2022 05:03:25 +0000 (+0530)
Subject: src/stop.sh: check if ceph-mds daemon(s) exist beforehand
X-Git-Tag: v18.0.0~928^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=8a4a432e8b6e41ff13c33698cc053079f6786aa2;p=ceph.git

src/stop.sh: check if ceph-mds daemon(s) exist beforehand

qa/tasks/vstart_runner.py runs src/stop.sh and src/vstart.sh if --create
is passed to it. Once in a few times, vstart_runner.py hangs on running
stop.sh. Running "ps -ef | grep ceph" shows that following command is
launched every time vstart_runner.py hangs at stop.sh -

  /usr/bin/python3.9 bin/ceph -c <path-to-ceph-repo>/build/ceph.conf tell mds.* client ls

Every time an instance of vstart_runner.py hangs at execution of
stop.sh, a new "ceph tell mds.* client ls" command is launched. This
doesn't happen when stop.sh is run manually. I suspect some issue lies
between this commmand in stop.sh and Python subprocess module.

Anyways, a simple fix is to not to run this command when ceph-mds daemon(s)
are not present on the system.

Signed-off-by: Rishabh Dave <ridave@redhat.com>
---

diff --git a/src/stop.sh b/src/stop.sh
index 71bc40986d10..60b6216b6840 100755
--- a/src/stop.sh
+++ b/src/stop.sh
@@ -104,8 +104,11 @@ do_umountall() {
     done
 
     #Get fuse mounts of the cluster
-    CEPH_FUSE_MNTS=$("${CEPH_BIN}"/ceph -c $conf_fn tell mds.* client ls 2>/dev/null | grep mount_point | tr -d '",' | awk '{print $2}')
-    [ -n "$CEPH_FUSE_MNTS" ] && sudo umount -f $CEPH_FUSE_MNTS
+    num_of_ceph_mdss=$(ps -e | grep \ ceph-mds$ | wc -l)
+    if test num_of_ceph_mdss -ne 0; then
+        CEPH_FUSE_MNTS=$("${CEPH_BIN}"/ceph -c $conf_fn tell mds.* client ls 2>/dev/null | grep mount_point | tr -d '",' | awk '{print $2}')
+        [ -n "$CEPH_FUSE_MNTS" ] && sudo umount -f $CEPH_FUSE_MNTS
+    fi
 }
 
 usage="usage: $0 [all] [mon] [mds] [osd] [rgw] [nfs] [--crimson] [--cephadm]\n"