cluster: STABLE3 - rgmanager: Do not fail service if non-critical resources fail to stop

Lon Hohberger lon at fedoraproject.org
Fri Sep 24 22:19:44 UTC 2010


Gitweb:        http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=7097cb971d5aa2d7bee6f2ac5bfd7e2db53c8f2d
Commit:        7097cb971d5aa2d7bee6f2ac5bfd7e2db53c8f2d
Parent:        d76d0935b4f130a93bf2b50f95b4a54a9264b4ab
Author:        Lon Hohberger <lhh at redhat.com>
AuthorDate:    Wed Sep 8 16:20:19 2010 -0400
Committer:     Lon Hohberger <lhh at redhat.com>
CommitterDate: Fri Sep 24 18:17:22 2010 -0400

rgmanager: Do not fail service if non-critical resources fail to stop

During non-critical component recovery (e.g. stop and leave dead),
rgmanager stops parts up to the __independent_subtree="2" flag.
If any stop operations fail during this phase, rgmanager should
ignore them since the resources are non-critical.

During a normal relocation/restart/enable/disable/etc, these
operations are critical and a fail to stop will result in the
service being placed in to the failed state.

Signed-off-by: Lon Hohberger <lhh at redhat.com>
---
 rgmanager/src/daemons/restree.c |   15 +++++++++++++++
 1 files changed, 15 insertions(+), 0 deletions(-)

diff --git a/rgmanager/src/daemons/restree.c b/rgmanager/src/daemons/restree.c
index ab0e20d..8719791 100644
--- a/rgmanager/src/daemons/restree.c
+++ b/rgmanager/src/daemons/restree.c
@@ -1510,6 +1510,21 @@ _res_op_internal(resource_node_t __attribute__ ((unused)) **tree,
 		node->rn_flags &= ~RF_NEEDSTOP;
 		rv |= res_exec(node, op, NULL, 0);
 
+		if (node->rn_flags & RF_QUIESCE) {
+			/* Non-critical resources = do not fail 
+			 * service if the resource fails to stop
+			 */
+			if (rv & SFL_FAILURE) {
+				logt_print(LOG_WARNING, "Failure to stop %s:%s"
+					   " during non-critical recovery "
+					   "operation\n",
+					   node->rn_resource->r_rule->rr_type,
+					   primary_attr_value(
+						node->rn_resource));
+				rv &= ~SFL_FAILURE;
+			}
+		}
+
 		if (rv == 0 && (node->rn_state == RES_STARTED ||
 				node->rn_state == RES_FAILED)) {
 			assert(node->rn_resource->r_incarnations >= 0);


More information about the cluster-commits mailing list