Rebuilding the ovdb

Peter Gradwell peter at
Mon Apr 30 15:41:58 UTC 2001


We had a case of expire crashing and never completing, and so
we've upgraded to -CURRENT as of Sunday to fix this. That seems
ok. We also had an overviewdb reporting the wrong information [1]
so we decided to rebuild it.

[ We're using buffindexed overview method. ]

- I started it off, having recreated it's buffers like so:
         makehistory -O -x -F

and nearly 24 hours later it's still going strong using inndf tells me
that it's doing a nice job of recreating the overviewdb.

Now, we only have 50Gig of nearly all CNFS news spool and it's
still going.

(a) it's taking forever. Last time I did this it took about 12 hours.
(b) is there anyway of finding out how much/many entries it
will create in the ovdb in total? - So we can guess how far it's
got to go.
(c) it's averaging only about 10% of CPU. Why wouldn't it use
more? Is it likely to be IO bound?



[1] e.g:

 >200 InterNetNews server INN 2.3.1 ready
>mode reader
>200 InterNetNews NNRP server INN 2.3.1 ready (posting ok).
>group uk.telecom
>211 4279 25144 33372 uk.telecom
>head 25144
>423 Bad article number

peter gradwell. gradwell dot com Ltd.
engineering & technical services for email, usenet and sms

More information about the inn-workers mailing list