You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Daniel Li <da...@gmail.com> on 2016/03/28 00:19:17 UTC

Exclude certain supervisors?

Due to some network switch issues, some supervisors are not stable
(connection on and off) in my production cluster. I'd like to temporarily
exclude those supervisors from connecting to the Storm cluster. Is there a
configuration change I can deploy to do so?

thx
Daniel

Re: Running cron jobs in a topology main?

Posted by Alexander T <mi...@gmail.com>.
Hello,

When you start Quartz it launches non-daemon backround threads which
prevents the JVM from exiting cleanly. Delay starting Quartz until your
spout is activated and stop it again in deactivate and you will be fine.
Otherwise you can set the "daemon" flags in Quartz to stop it from blocking
JVM exit.

We ditched Quartz and used Springs CronSequenceGenarator instead. This fits
better with Storms polling architecture and doesn't requiring starting any
threads at all.

Best regards
Alexander
On Mar 28, 2016 07:08, "이승진" <sw...@navercorp.com> wrote:

>
>
> Hello all,
>
>
>
> If I were to run some cron jobs inside of topology, is it proper to
> declare them in the topology main method?
>
>
>
> Previously for that purpose I made custom spout inside of which runs a
> cron scheduler, set the parallelism of it to 1, and it's just working fine.
>
>
>
> If that cron declaration part is in topology main, I think topology
> submitting process never ends and remain in the nimbus server without being
> managed.
>
>
>
> Do you have any opinion on this topic? I'm using 0.9.x version of storm by
> the way.
>
>
>
> Thanks in advance.
>

Running cron jobs in a topology main?

Posted by 이승진 <sw...@navercorp.com>.
  
Hello all,
 
If I were to run some cron jobs inside of topology, is it proper to declare them in the topology main method?
 
Previously for that purpose I made custom spout inside of which runs a cron scheduler, set the parallelism of it to 1, and it's just working fine.
 
If that cron declaration part is in topology main, I think topology submitting process never ends and remain in the nimbus server without being managed.
 
Do you have any opinion on this topic? I'm using 0.9.x version of storm by the way.
 
Thanks in advance. 




Re: 答复: Exclude certain supervisors?

Posted by Salimane Adjao Moustapha <me...@salimane.com>.
Hi,

@Daniel Li, @John Fang, it could be awesome to include an example in the
storm readme

Thanks

*SALIMANE ADJAO MOUSTAPHA*


On Mon, Mar 28, 2016 at 7:19 AM, John Fang <xi...@alibaba-inc.com>
wrote:

> Yes, we have health check for supervisor. You can put the script which
> check the connections in the directory  “storm.health.check.dir”, we will
> check the supervisor’s healthy status by the script at regular intervals.
> Once we check the supervisor is unhealthy, then we will kill the supervisor
> and his workers. But you need provide the script.  And the script must
> output “ERROR” when it find the supervisor war unhealthy.
>
>  You  can refer to “healthcheck.clj” before you write  the script.
>
>
>
> Regards
>
> John Fang
>
>
>
>
>
> *发件人:* Daniel Li [mailto:danielli90@gmail.com]
> *发送时间:* 2016年3月28日 6:19
> *收件人:* user@storm.apache.org
> *主题:* Exclude certain supervisors?
>
>
>
> Due to some network switch issues, some supervisors are not stable
> (connection on and off) in my production cluster. I'd like to temporarily
> exclude those supervisors from connecting to the Storm cluster. Is there a
> configuration change I can deploy to do so?
>
>
>
> thx
>
> Daniel
>

答复: Exclude certain supervisors?

Posted by John Fang <xi...@alibaba-inc.com>.
Yes, we have health check for supervisor. You can put the script which check the connections in the directory  “storm.health.check.dir”, we will check the supervisor’s healthy status by the script at regular intervals. Once we check the supervisor is unhealthy, then we will kill the supervisor and his workers. But you need provide the script.  And the script must output “ERROR” when it find the supervisor war unhealthy. 

 You  can refer to “healthcheck.clj” before you write  the script.

 

Regards

John Fang

 

 

发件人: Daniel Li [mailto:danielli90@gmail.com] 
发送时间: 2016年3月28日 6:19
收件人: user@storm.apache.org
主题: Exclude certain supervisors?

 

Due to some network switch issues, some supervisors are not stable (connection on and off) in my production cluster. I'd like to temporarily exclude those supervisors from connecting to the Storm cluster. Is there a configuration change I can deploy to do so?

 

thx

Daniel