[request-tracker-maintainers] Bug#623553: Bug#623553: additionnal details

[Dominic Hargreaves]

On Thu, Apr 21, 2011 at 12:42:34PM -0400, anarcat wrote:
> Oh and note that in our case, it's not a "sudden jump in usage, not a
> gradual leak" - the leak is progressive and *will* eat up all memory on
> the server eventually. After 30 seconds (the fastcgi timeout), the
> memory usage is around 1GB, iirc.

Yes; bit the leak is rapid (as you've confirmed above) and only happens
in very specific circumstances - that's what I meant by a sudden jump.
Measured across many requests, it is.

> I am not sure it's FastCGI's responsability to cleanup that stuff -
> Apache connects to the fastcgi threads, which are usually long-lived. If
> they timeout once, that's not a good reason to kill them...
> 
> I am not sure the solution proposed in #15108 is a good one as it would
> hide the problem in this case.

Sure; it's a workaround, not a fix, and that is the spirit in which it
was submitted. Without having a diagnosis, it was (and still is)
useful.

Dominic.

-- 
Dominic Hargreaves | http://www.larted.org.uk/~dom/
PGP key 5178E2A5 from the.earth.li (keyserver,web,email)