[Sheepdog] [PATCH] sheep: fix recovery logic

Liu Yuan namei.unix at gmail.com
Mon Feb 27 07:46:10 CET 2012


On 02/23/2012 06:10 PM, Liu Yuan wrote:

> From: Liu Yuan <tailai.ly at taobao.com>
> 
> We should check if join_msg->inc_epoch before starting recovery after
> a new jode joins.
> 
> Signed-off-by: Liu Yuan <tailai.ly at taobao.com>
> ---
>  sheep/group.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/sheep/group.c b/sheep/group.c
> index e81007c..d10ba06 100644
> --- a/sheep/group.c
> +++ b/sheep/group.c
> @@ -820,7 +820,7 @@ static void __sd_join_done(struct cpg_event *cevent)
>  
>  	update_cluster_info(jm, &w->joined, w->member_list, w->member_list_entries);
>  
> -	if (sys_can_recover()) {
> +	if (sys_can_recover() && jm->inc_epoch) {
>  		list_for_each_entry_safe(node, t, &sys->leave_list, list) {
>  			list_del(&node->list);
>  		}


Applied to master.

Thanks,
Yuan



More information about the sheepdog mailing list