Hi,
> FYI, the "ocf" drbd agent does loop on monitor,
> and should return a generic error in that case already.
That's good!
> Maybe we should add such a loop to drbddisk as well.
> Or somehow set it up as a wrapper around the ocf agent (though that may
> not be easily possible).
>
> Yes, your patch is ok.
> Still I'm not taking it as such, but probably make drbdsetup more robust
> in face of file system problems on /var/lock/, and add a monitoring loop
> to drbddisk instead.
make sense.
I modified my patch.
It imitates drbd_stop() in ocf drbd.
Thanks,
Junko
--- drbddisk.org 2010-06-15 17:48:30.000000000 +0900
+++ drbddisk.tmp 2010-07-12 18:44:43.000000000 +0900
@@ -44,28 +44,21 @@ case "$CMD" in
done
;;
stop)
- $DRBDADM secondary $RES
- ex=$?
- case $ex in
- 0)
- exit 0
- ;;
- 11)
- # see drbdadm_main.c adm_generic and m_system
- # as well as drbdsetup.c:
- # in fact a role change was attempted, but failed.
- echo >&2 "$DRBDADM secondary $RES: exit code $ex, mapping to 1"
- exit 1 # LSB generic error
- ;;
- *)
- # other error, may be syntax error in config file,
- # anything else: to not confuse heartbeat further,
- # and avoid reboot due so "failed stop recovery",
- # pretend that we succeeded in stopping this.
- echo >&2 "$DRBDADM secondary $RES: exit code $ex, mapping to 0"
- exit 0
- ;;
- esac
+ rc=1
+ while true; do
+ $DRBDADM secondary $RES
+ status=$?
+ case $status in
+ 0)
+ rc=0
+ break
+ ;;
+ *)
+ sleep 1
+ ;;
+ esac
+ done
+ exit $rc
;;
status)
if [ "$RES" = "all" ]; then
_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user
