If failed_to_recv is set (node detect itself not able to receive
message), we can end up with assert, because my_failed_list and
my_member_list are same list. This is happening because we are not
following specification and we allow to mark node itself as failed.
Because if failed_to_recv is set and we reached consensus across nodes,
single node membership is created (ignoring both fail list and
member_list), we can skip assert.
Signed-off-by: Jan Friesse <jfriesse@xxxxxxxxxx>
---
exec/totemsrp.c | 15 +++++++++++++++
1 files changed, 15 insertions(+), 0 deletions(-)
diff --git a/exec/totemsrp.c b/exec/totemsrp.c
index ec951df..a4cc19a 100644
--- a/exec/totemsrp.c
+++ b/exec/totemsrp.c
@@ -1247,6 +1247,16 @@ static int memb_consensus_agreed (
break;
}
}
+
+ if (agreed && instance->failed_to_recv == 1) {
+ /*
+ * Both nodes agreed on our failure. We don't care how many proc list items left because we
+ * will create single ring anyway.
+ */
+
+ return (agreed);
+ }
+
assert (token_memb_entries >= 1);
return (agreed);
@@ -3617,6 +3627,11 @@ printf ("token seq %d\n", token->seq);
instance->my_aru_count = 0;
}
+ /*
+ * We really don't follow specification there. In specification, OTHER nodes
+ * detect failure of one node (based on aru_count) and my_id IS NEVER added
+ * to failed list (so node never mark itself as failed)
+ */
if (instance->my_aru_count > instance->totem_config->fail_to_recv_const &&
token->aru_addr == instance->my_id.addr[0].nodeid) {
--
1.7.1
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss
