This patch passes two test cases:
-------
Test #1
-------
Two node cluster - run cpgbench on each node
modify totemsrp with following defines:
Two test cases:
-------
Test #2
-------
5 node cluster
start 5 nodes randomly at about same time, start 5 nodes randomly at about
same time, wait 10 seconds and attempt to send a message. If message blocks
on "TRY_AGAIN" likely a message loss has occured. Wait a few minutes without
cyclng the nodes and see if the TRY_AGAIN state becomes unblocked.
If it doesn't the test case has failed
Signed-off-by: Steven Dake <sdake at redhat.com>
---
exec/totemsrp.c | 31 ++++++++++++++++++++++++++++++-
1 files changed, 30 insertions(+), 1 deletions(-)
diff --git a/exec/totemsrp.c b/exec/totemsrp.c
index 6981ac1..5b61f3b 100644
--- a/exec/totemsrp.c
+++ b/exec/totemsrp.c
@@ -1794,7 +1794,36 @@ static void memb_state_operational_enter (struct totemsrp_instance *instance)
sizeof (struct srp_addr) * instance->my_memb_entries);
instance->my_failed_list_entries = 0;
- instance->my_high_delivered = instance->my_high_seq_received;
+ /*
+ * TODO Not exactly to spec
+ *
+ * At the entry to this function all messages without a gap are
+ * deliered.
+ *
+ * This code throw away messages from the last gap in the sort queue
+ * to my_high_seq_received
+ *
+ * What should really happen is we should deliver all messages up to
+ * a gap, then delier the transitional configuration, then deliver
+ * the messages between the first gap and my_high_seq_received, then
+ * deliver a regular configuration, then deliver the regular
+ * configuration
+ *
+ * Unfortunately totempg doesn't appear to like this operating mode
+ * which needs more inspection
+ */
+ i = instance->my_high_seq_received + 1;
+ do {
+ void *ptr;
+
+ i -= 1;
+ res = sq_item_get (&instance->regular_sort_queue, i, &ptr);
+ if (i == 0) {
+ break;
+ }
+ } while (res);
+
+ instance->my_high_delivered = i;
for (i = 0; i <= instance->my_high_delivered; i++) {
void *ptr;
--
1.7.6.2
