[sheepdog] [PATCH] zookeeper: add more detailed description on how zk_watcher report states

Liu Yuan namei.unix at gmail.com
Thu Feb 26 04:40:35 CET 2015


On Wed, Feb 11, 2015 at 06:42:16PM +0800, Liu Yuan wrote:
> From: Liu Yuan <liuyuan at cmss.chinamobile.com>
> 
> Signed-off-by: Liu Yuan <liuyuan at cmss.chinamobile.com>
> ---
>  sheep/cluster/zookeeper.c | 25 +++++++++++++++++++++++--
>  1 file changed, 23 insertions(+), 2 deletions(-)
> 
> diff --git a/sheep/cluster/zookeeper.c b/sheep/cluster/zookeeper.c
> index b603d36..e2ee248 100644
> --- a/sheep/cluster/zookeeper.c
> +++ b/sheep/cluster/zookeeper.c
> @@ -685,6 +685,27 @@ static int add_event(enum zk_event_type type, struct zk_node *znode, void *buf,
>  	}
>  }
>  
> +/*
> + * Type value:
> + * -1 SESSION_EVENT, use State to indicate what kind of sub-event
> + *   State value:
> + *   -122 SESSION EXPIRED
> + *   1    CONNECTING
> + *   3    CONNECTED
> + * 1  CREATED_EVENT
> + * 2  DELETED_EVENT
> + * 3  CHANGED_EVENT
> + * 4  CHILD_EVENT
> + *
> + * While connection to zk is disconnected (zk cluster in election, network is
> + * broken, etc.), zk library will try to reconnect zk cluster on its own and
> + * report both the connection state and session state to sheep via zk_watcher.
> + *
> + * Once the connection is reestablished (state changed from 1 to 3)within the
> + * timeout window, the session is still valid, meaning that all the watchers
> + * will function as before. If not within the timeout window, zk_watcher will
> + * report to sheep that session is expired.
> + */
>  static void zk_watcher(zhandle_t *zh, int type, int state, const char *path,
>  		       void *ctx)
>  {
> @@ -693,6 +714,8 @@ static void zk_watcher(zhandle_t *zh, int type, int state, const char *path,
>  	uint64_t lock_id;
>  	int ret;
>  
> +	sd_debug("path:%s, type:%d, state:%d", path, type, state);
> +
>  	if (type == ZOO_SESSION_EVENT && state == ZOO_EXPIRED_SESSION_STATE) {
>  		/*
>  		 * do reconnect in main thread to avoid on-the-fly zookeeper
> @@ -702,8 +725,6 @@ static void zk_watcher(zhandle_t *zh, int type, int state, const char *path,
>  		return;
>  	}
>  
> -/* CREATED_EVENT 1, DELETED_EVENT 2, CHANGED_EVENT 3, CHILD_EVENT 4 */
> -	sd_debug("path:%s, type:%d, state:%d", path, type, state);
>  	if (type == ZOO_CREATED_EVENT || type == ZOO_CHANGED_EVENT) {
>  		ret = sscanf(path, MEMBER_ZNODE "/%s", str);
>  		if (ret == 1)
> -- 
> 1.9.1
> 

Applied since it is a cosmetical patch, no functional change.

Yuan



More information about the sheepdog mailing list