Hi,

Am 20.03.2013 um 13:01 schrieb Txema Heredia Genestar:

> We have a 300-core cluster. Our users have always submitted their jobs using 
> "-m ea" to receive an email whenever the job finishes or aborts. Our typical 
> user tends to submit 1.000~3.000 jobs at once. They usually don't use task 
> jobs, but submit each one as an independent job. In addition, we have an 
> epilog script that, whenever a job finishes, checks the requested memory 
> (h_vmem) and the real maxvmem used, and sends an email to the user if they 
> are blatantly requesting more than they need.
> This leads that, on a typical "run", our users receive from 1.000 to 6.000 
> emails from the cluster.
> 
> Everybody was ok with this but, a few months ago, our university moved the 
> mail service to google and now the mail admins are threatening us to ban all 
> mails from our cluster.
> 
> Is there any solution to this (beyond not sending mails or changing the 
> task-job habits of our users)? Are there any addons/plugins out there that 
> accumulate mails in the master node and send hourly/daily digests to the 
> users?

This was on the list before some times, to get only one email when the complete 
array job finished in total. What you can implement:

a) Don't send emails from the array job, but submit a dummy job afterwards with 
-hold_jid of the array job (it could even run in a dummy-queue on the headnode 
solely for this purpose as it won't put any load on the machine), which sends 
just one eMail after the main job. It could also perform a `qacct` to extract 
the exit_codes from all the array instances.

b) Most likely you have an MTA agent running on the master node of the cluster 
to forward it to Google right now. It should also be possible to use a 
regexp/hash recipient_canonical to mangle the official email address like 
[email protected] into foobar@headnode and collect them there first. A cron 
job could then assemble one email per hour which includes all the emails for 
this particular user of the last hour and send this one to Google.

-- Reuti


> Thanks in advance,
> 
> Txema
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to