Have you tried Dave's cpg test tool? Would be interesting to see if
this introduces any other regressions.
Reviewed-by: Steven Dake <sdake@xxxxxxxxxx>
On 06/13/2012 07:07 AM, Jan Friesse wrote:
> Test scenario is follows:
> - node 1, node 2
> - node 1 is paused
> - node 2 sees node 1 dead
> - node 1 unpaused
> - node 1 and 2 both choose same dowlist message which includes node 2 ->
> node 2 is efectivelly disconnected
>
> Patch includes additional test if left_node is localnode. If so, such
> downlist is ignored.
>
> Signed-off-by: Jan Friesse <jfriesse@xxxxxxxxxx>
> (backported to flatiron)
> ---
> services/cpg.c | 21 +++++++++++++++++++++
> 1 files changed, 21 insertions(+), 0 deletions(-)
>
> diff --git a/services/cpg.c b/services/cpg.c
> index d7a26f2..dbd61c5 100644
> --- a/services/cpg.c
> +++ b/services/cpg.c
> @@ -729,6 +729,8 @@ static struct downlist_msg* downlist_master_choose (void)
> struct list_head *iter;
> uint32_t cmp_members;
> uint32_t best_members;
> + uint32_t i;
> + int ignore_msg;
>
> for (iter = downlist_messages_head.next;
> iter != &downlist_messages_head;
> @@ -736,10 +738,26 @@ static struct downlist_msg* downlist_master_choose (void)
>
> cmp = list_entry(iter, struct downlist_msg, list);
> downlist_log(LOGSYS_LEVEL_DEBUG, "comparing", cmp);
> +
> + ignore_msg = 0;
> + for (i = 0; i < cmp->left_nodes; i++) {
> + if (cmp->nodeids[i] == api->totem_nodeid_get()) {
> + log_printf (LOG_DEBUG, "Ignoring this entry because I'm in the left list\n");
> +
> + ignore_msg = 1;
> + break;
> + }
> + }
> +
> + if (ignore_msg) {
> + continue ;
> + }
> +
> if (best == NULL) {
> best = cmp;
> continue;
> }
> +
> best_members = best->old_members - best->left_nodes;
> cmp_members = cmp->old_members - cmp->left_nodes;
>
> @@ -754,6 +772,9 @@ static struct downlist_msg* downlist_master_choose (void)
> }
>
> }
> +
> + assert (best != NULL);
> +
> return best;
> }
>
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss
