inn2.4.1, expire running > 2 days on a feeder

[Chris Stromsoe]

On Thu, 23 Sep 2004, Miquel van Smoorenburg wrote:
> In article <Pine.LNX.4.61.0409230256180.11485 at potato.cts.ucla.edu>, 
> Chris Stromsoe <cbs at cts.ucla.edu> wrote:
>> Starting about a week ago, expire started running for several days at a 
>> time.  It used to run for 6 or 7 hours.  At some point, history crossed 
>> the 2gb boundary.  That may or may not have anything to do with it. 
>> inn was built with --enable-largefiles.  The machine is running debian 
>> "testing" linux.
>>
>> spool is cnfs, and overview is disabled.  news is stored in 3 x 10Gb 
>> buffers, 1 x 40Gb buffer, and 1 x 5Gb buffer, on a single jfs partition 
>> built from a 4 disk scsi raid5 with no spare.  ~news/db/ is on the same 
>> partition.
>>
>> expire was killing performance, so I killed it.  Any ideas about what 
>> to look for to figure out what's wrong?
>
> You probably need more memory. I/O from incoming/outgoing feeds is 
> throwing away the existing cached history file, etc.

Hrm.  I'll try adding some.  Before I go nuts with ram, how much are you 
thinking?  The box is a dual xeon with 1Gb.  One of my readers is working 
with a 25Gb history file and the same amount or memory, but only runs for 
around 3 hours to complete news.daily and also has 1Gb (and slower cpus).

> On my system I simply throttle INN before expire and start it after 
> expire. news.daily actually has some support for this though you have to 
> read and grok the script to be able to use it.
>
> If you set the MESSAGE environment variable, the script will throttle 
> innd before expire. You need to pass -n as well. So, my crontab looks 
> like:

I'll try that and see if it helps.


-Chris