Expireover running really slow

[Kirill Berezin]

Marc Stürmer wrote:
> Am Mittwoch 09 Januar 2008 schrieben Sie:
>
>   
>> This is a real problem for running innd with large storages. Performance
>> degrades because overview records are highly scattered over big area,
>> ovdb does not help. An expiration for ovdb works much faster for first
>>     
>
> OVDB is the worst of the backends in time of expiration. Buffindexed may be 
> somewhat faster, but it lost too many articles for me when I used it on my 
> machine, so it's also not really an option for me anymore. 
>
> The best of the overview-backends for me is really tradindexed now. Works 
> good, no performance loss compared to ovdb and expiration time shrank from 6 
> hours under ovdb to 1 hour. 
>   
How many articles are to be removed every day for you server? Mean 
value, if possible, please, just for statistics.
> Also please note, that buffindexed is not a circular system like CNFS, meaning 
> when your buffers are full, they're full and you got a problem, because 
> overview will just cease to work then at all, so you always should have added 
> some spare buffers "just in case" in your config if that case happens. 
>   
Expiryover frees a space removing outdated overview records. If you have a reasonable expiration policy, then overflow never happen.

kirill