https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6075





--- Comment #29 from AXB <[email protected]>  2009-03-03 22:24:55 PST ---
(In reply to comment #28)
> > No problem, and in fact thanks for sticking with it this long. i'm very 
> > happy
> > to have documented in the record these difficult to debug problems that 
> > could
> > arise from a high speed connection throttled by a lower speed upstream, or 
> > by
> > high rate UDP getting throttled by a Zyxel or similarly performing router.
> 
> I wonder how common this problem is. Thanks to Elsa for perseverance!
> How many SOHO users of SpamAssassin suffer from this without knowing?

Please note :

This has been noticed/debugged in a SOHO setup with unusually high traffic
(+80 simultaneous SMTP sessions) doing inline scanning and not using sequential
lookups as would happen with a MDA. (specifically: spam trap boxes)

Hardware was +5 year old and not ready for today's high speed connections.
Updated hardware does not show this effect.
In one case, where a custoemr was affected, reporting to Zyxel, resulted in a
firmware update which corrected the issue.

> It might be possible to implement application-level traffic shaping
> (leaky bucket or token bucket algorithm) for taming UDP storms.
> The main problem seems to be lack of fine granularity in tasks of SA
> (and the absence of fine granularity interrupts in Perl). Something
> like a 10 ms scheduling event would do for the purpose, and could also
> help with more prompt harvesting of responses (Bug 5589 comment 7).
> A Bug 6060 calls for shorter chunks of noninterrupted grinding too.
> Just thinking aloud...

IMO, adding throttling won't fix the issue as timeframes may vary in such a
wide way that it would only make debugging harder.


-- 
Configure bugmail: 
https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

Reply via email to