> Sorry if this is already answered in prior threads.  I did a search
> backward in the archives for the last few months, but didn't see an
> answer.
>
> We have uWSGI in place running a pyramid stack behind nginx in turn behind
> a load balancer with 2 virtual servers running a fairly low load service,
> but it will scale up over time.  We would like to set it up to do 2
> things:
> 1. deploy new code to a single server for initial monitoring/live
> validation similar to the Zerg testing mode described in the docs (
> http://uwsgi-docs.readthedocs.org/en/latest/Zerg.html)
> 2. gracefully deploy the new code and restart all servers with the new
> code.
>
> #2 is hinted at in the docs and in prior emails, but no recipe is written
> out that I can find.  I'd love to see how you guys intended that to get
> set
> up.
>
> From the docs I think setting the system up for auto-scaling
> (emperor+brrodlord+zerg) along with SIGTSTP for #1 and SIGHUP for #2 would
> do the trick.  Question is, will the emperor fire up the SOS Zerg
> processes
> when we issue those signals or will it not because no vassal is sending
> message 30 to the emperor?
>
>

I do not think broodlord mode will be useful here.

I think the easiest way would be having a persistent zergpool (it is an
instance running to pass the socket to zergs) and simply suspend/resume
the various instances.

Zergpools are described at the end of the zerg page (zergpools instances
can be managed by the emperor too)


-- 
Roberto De Ioris
http://unbit.it
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi

Reply via email to