Eventscripts: change failure counts and behaviour for statd and nfsd.

We reduce the number of failures before attempting a restart.
However, after 6 failures we mark the cluster unhealthy and no longer
try to restart.  If the previous 2 attempts didn't work then there
isn't any use in bogging the system down with an attempted restart on
every monitor event.

Signed-off-by: Martin Schwenke <martin@meltin.net>
(This used to be ctdb commit f654739080b40b7ac1b7f998cacc689d3d4e3193)

diff --git a/ctdb/config/events.d/60.nfs b/ctdb/config/events.d/60.nfs
index 43d5f665dbb30dd0f7e67aa6a3bcb3583c6d6e18..2b4c158d6835c98864966d65e976eed7808bd78c 100755 (executable)
--- a/ctdb/config/events.d/60.nfs
+++ b/ctdb/config/events.d/60.nfs
@@ -76,13 +76,16 @@ case "$1" in
         p="rpc.statd"
         which $p >/dev/null 2>/dev/null && \
            nfs_check_rpc_service "statd" \
-           -ge 10 "verbose restart"
+               -ge 6 "verbose unhealthy" \
+               -eq 4 "verbose restart" \
+               -eq 2 "restart:bs"
 
        # check that NFS responds to rpc requests
        if [ "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" != "yes" ] ; then
            nfs_check_rpc_service "knfsd" \
-               -ge 15 "verbose restart unhealthy" \
-               -eq 10 "restart:bs"
+               -ge 6 "verbose unhealthy" \
+               -eq 4 "verbose restart" \
+               -eq 2 "restart:bs"
        fi
 
        # check that lockd responds to rpc requests