Tony Finch wrote:
> On Wed, 29 Jun 2005, Mark Lowes wrote:
>
>>The actual error I was encountering was a "timeout getting lock",
>>causing slowdown on handling incoming connections before it gave up
>>trying to work out the rate limiting stuff. It's possible that my
>>ratelimiting rules are a little aggressive which isn't helping the
>>situation but as traffic ramps up the problem would have been reached
>>anyway.
>
>
> You are using a *lot* of ratelimit rules, which implies thrashing the
> ratelimit hints db lock. Although the code caches the results of ratelimit
> calculations where appropriate, you can defeat this by using per_rcpt
> limits etc. Perhaps I should add a caveat to the documentation.
Well there's always some sod who pushes the code beyond what was
initially expected. I'm still tuning the limits and I can probably
start looking at my logs to see which rules are actually getting
tripping and which aren't and use that as the first stage of optimisation.
Mark
