[sheepdog] [PATCH] sheep/zookeeper: ensure zookeeper is connected

[Ruoyu]

On 2014年07月23日 12:19, Hitoshi Mitake wrote:
> At Wed, 23 Jul 2014 12:02:10 +0800,
> Ruoyu wrote:
>> zookeeper session establishment is asynchronous. It means that
>> we cannot judge the session is connected by checking the
>> return value of the func zookeeper_init. Instead, we should
>> check the zookeeper handler's state in a loop to ensure
>> the session is connected.
>>
>> Signed-off-by: Ruoyu <liangry at ucweb.com>
>> ---
>>   sheep/cluster/zookeeper.c | 16 ++++++++++++++--
>>   1 file changed, 14 insertions(+), 2 deletions(-)
> Applied, thanks.
>
> It would be nice if the value of interval can be specified by user.
Maybe keep it simple is better. If I am the user, I don't care how many 
times sheep tries to connect zookeeper cluster. I just care whether the 
session is connected or not during expiry.
>
> Thanks,
> Hitoshi
>
>> diff --git a/sheep/cluster/zookeeper.c b/sheep/cluster/zookeeper.c
>> index 64613ca..96acce6 100644
>> --- a/sheep/cluster/zookeeper.c
>> +++ b/sheep/cluster/zookeeper.c
>> @@ -1351,7 +1351,7 @@ static void zk_unlock(uint64_t lock_id)
>>   static int zk_init(const char *option)
>>   {
>>   	char *hosts, *to, *p;
>> -	int ret;
>> +	int ret, interval, retry = 0, max_retry;
>>   
>>   	if (!option) {
>>   		sd_err("You must specify zookeeper servers.");
>> @@ -1371,10 +1371,22 @@ static int zk_init(const char *option)
>>   		 ZOO_MINOR_VERSION, ZOO_PATCH_VERSION, hosts, zk_timeout);
>>   	zhandle = zookeeper_init(hosts, zk_watcher, zk_timeout, NULL, NULL, 0);
>>   	if (!zhandle) {
>> -		sd_err("failed to connect to zk server %s", option);
>> +		sd_err("failed to initialize zk server %s", option);
>>   		return -1;
>>   	}
>>   
>> +	/* the simplest way to wait and check zk connection */
>> +	interval = 100;
>> +	max_retry = zk_timeout / interval;
>> +	while (zoo_state(zhandle) != ZOO_CONNECTED_STATE) {
>> +		usleep(interval * 1000);
>> +		if (++retry >= max_retry) {
>> +			sd_err("failed to connect to zk server %s "
>> +					"after %d retries", option, retry);
>> +			return -1;
>> +		}
>> +	}
>> +
>>   	uatomic_set_false(&stop);
>>   	uatomic_set_false(&is_master);
>>   	if (zk_queue_init() != ZOK)
>> -- 
>> 1.8.3.2
>>
>>
>> -- 
>> sheepdog mailing list
>> sheepdog at lists.wpkg.org
>> http://lists.wpkg.org/mailman/listinfo/sheepdog