2011/9/14 Roberto De Ioris <[email protected]>:
> This is a warn for your configuration not being able to manage the current
> rate of requests.

But few seconds before everything is fine (load avg ~2.0-3.0). And
then bang, load avg. 30-40, and site is down. Restart - again,
everything is fine, same traffic.

> Your 48 processes scares me a bit, on high-load you will ends with 48
> postgresql and 48 uwsgi processes probably devouring your ram/cpu.

We are using pgbouncer. Every service is on dedicated server.

There is server with nginx and static files, two servers with uWSGI
(and only uWSGI, 8 GB of RAM each) and DB server (16 GB of ram,
8-disks RAID10)

> (try to launch a top when this happens, to see the machine status)

One strange thing is "flush-0:21" on uWSGI workers using 99% of CPU.
As far I know it's kernel garbage collector?

>
> I suggest you decreasing them (there is no one-fit-for-all value).

Ok. Decreasing to 32... we will see. :)

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

Reply via email to