Re: [exim] High Perf server - was (exim allowed someone to s…

Startseite
Nachricht löschen
Nachricht beantworten
Autor: Mark Lowes
Datum:  
To: andrew.johnson
CC: exim-users, peter
Betreff: Re: [exim] High Perf server - was (exim allowed someone to slam my mailserver for 3 hours)
Andrew Johnson wrote:
>>Exim doesn't support tightly-coupled clusters. This is a
>>weakness in the implementation of the hints DBs, and fixing
>>it was outside the scope of the ratelimit project. There are
>>other things that would benefit from better hints DBs: the
>>current implementation is a major bottleneck for high
>>capacity servers, so there's scope for improved performance;
>>sharing the hints DBs will reduce duplicated work across a
>>cluster (retrys, callouts, etc.); greylisting would be much
>>better if implemented cluster-wide rathe than per-machine.
>>Perhaps one day someone will tackle it :-)
> Any thoughts on how to improve the current performance of the hints
> databases ?.


Mount /var/spool/exim/db/ as a ramdisk.

> multiple load balanced servers) but I have the same concerns about the db's
> - particually as now the rate limit stuff uses them too.


We hit a bottleneck in db access when I rolled out the ratelimiting,
moving to a ramdisk has resolved the problem for the moment, though I'm
not entirely sure what the next step will be.

    Mark