Pulse web server connection limit questions

[nrusch]

I’m wondering if you guys have any opinions on what a “sane” connection limit for the Pulse web service would be. The default setting of 10 seems extremely low, and I don’t know if this is out of a concern for performance, or the result of something else.

We are performing all of our job submissions, as well as a lot of cron-like farm housekeeping operations, using the REST API, and I’d like to make sure we don’t run into connection issues. I’ve got the limit set at 35 right now, but I’d like to at least double it if possible.

Also, what happens if a client attempts to connect while the connection capacity is at maximum? Will it just refuse the connection at the socket level, or does it actually allow the connection, but respond immediately with some sort of error status?

Thanks for any info.

[rrussell]

We added the web service to Pulse back when it played a major role in job scheduling, and to play it safe, we defaulted it to 10 so that it wouldn’t impact the serving of jobs. When we took the job scheduling functionality out in Deadline 6, we didn’t even think to bump up the default web service connection limit. Pulse was able to keep up with hundreds of slaves to serve them jobs, so a limit of 100 should still be playing it rather safe.

If a client attempts to connect while the connection limit is maxed out, the client will just wait until either a connection becomes free, or a timeout occurs.

Cheers,
Ryan

[nrusch]

Good to hear. Thanks Ryan.