ovdb expireover slow and a resource hog

[Helmut Kreft]

Hello,

I am currently working on a new server to replace our somewhat aged
INN-NiD 1.7.2. 

This machine has currently one text-only feed from our old machine.
What concerns me is this:

expire begin Fri Mar 15 01:01:15 MET 2002: (-v1 -z/srv/news/spool/log/expire.rm)
    Article lines processed  4279781
    Articles retained        3963091
    Entries expired           316690
    Articles dropped          316690
    Old entries dropped       337187
    Old entries retained      247183
expire end Fri Mar 15 01:03:24 MET 2002
        all done Fri Mar 15 01:03:24 MET 2002
        expirerm start Fri Mar 15 01:03:30 MET 2002
        expirerm end Fri Mar 15 01:04:07 MET 2002
        expireover start Fri Mar 15 01:04:07 MET 2002
        expireover end Fri Mar 15 02:09:49 MET 2002

Expireover is well over one hour. While expireover runs, the spool-
metadevice goes up from basically idle to 92% usage on average 
acording to sar and transfering around 2MB/s. I assume the drives 
have to seek *a lot*. 

01:04:40   device        %busy   avque   r+w/s  blks/s  avwait  avserv

Average:
...
           md9              93    21.1     248    4382   0.4    84.7
...

Is that normal behaviour for an expirover? 

System:
- SunOS abc.def.de 5.8 Generic_108528-13 sun4u sparc SUNW,Ultra-4
- INN 2.3.2, Berkeley DB 3.3.11, all timecaf, ovdb, cleanfeed
- E450, 3*400 MHZ UltraII, 1.25GB memory.
- Spool area for news consists of 4*36GB Fujitsu MAN3367MC drives in a 
  mirrored stripe setup (2 stripes both part of a mirror). Drives are
  evenly distributed over all SCSI-channels. 

    Helmut

-- 
In a world increasingly dominated by image culture (TV, movies, .jpg 
files), UNIX remains rooted in the culture of the word.
                                                    Thomas Scoville
-- 
In a world increasingly dominated by image culture (TV, movies, .jpg 
files), UNIX remains rooted in the culture of the word.
                                                    Thomas Scoville