At Thu, 30 May 2013 22:33:23 +0800, Liu Yuan wrote: > > On 05/30/2013 10:06 PM, MORITA Kazutaka wrote: > > At Wed, 29 May 2013 21:00:49 +0800, > > Liu Yuan wrote: > >> > >> On 05/29/2013 06:37 PM, MORITA Kazutaka wrote: > >>> + * Note that this function can return ZOPERATIONTIMEOUT or ZCONNECTIONLOSS and > >>> + * the caller has to handle it. > >>> */ > >> > >> It seems that zookeeper(v1.4.5) or upwards ( I didn't check it) return > >> > >> ZINVALIDSTATE - zhandle state is either ZOO_SESSION_EXPIRED_STATE or > >> ZOO_AUTH_FAILED_STATE > >> > >> for zoo_create. > > > > Yes. The comment might be misleading. What I wanted to write is that > > the caller has to retry zk_create_seq_node() when the error is > > retryable. I'll update the comment and send as v2. > > > > What I meant was, ZOPERATIONTIMEOUT, ZCONNECTIONLOSS are not returned by > zk_create. But actually I've experienced many ZCONNECTIONLOSS errors by zoo_create() with zookeeper 3.3.5. I'm not sure ZOPERATIONTIMEOUT also happens, though. Thanks, Kazutaka |