Bug in buffindexed?

The Hermit Hacker scrappy at hub.org
Sun Oct 3 19:42:22 UTC 1999


On Sun, 3 Oct 1999, Katsuhiro Kondou wrote:

> In article <Pine.BSF.4.10.9910021050390.485-100000 at thelab.hub.org>,
> 	The Hermit Hacker <scrappy at hub.org> wrote;
> 
> } Just ran out of overview space again on my system, so created a new buffer
> } and restarted...got an error to the effect of:
> } 
> } buffindexed: length mismatch '1572864000' for index '0' (1573164095 bytes)
> } 
> } So did an ls and check of my buffindexed.conf file:
> :
> } Aren't OV2 and OV3 just a little large?  When I created them, they were the
> } size I stipulated in buffindexed...
> 
> Hm, I hadn't seen that problem on my testing server,
> when its buffer was filled up.  I'll test again.
> You need to recreate whole overview anyway.

'K, that's what I ended up doing...

One thing I'm curious about, and that is how painfully slow it is to
recreate the overviews...rebuilding history takes no time, but I ran
'makehistory -I -O' on a PII server with 768Meg of RAM and something like
36gig of full CNFS buffers, and after ~14hrs I said to hell with it, wip'd
out the CNFS/overview buffers and started with empty spools again...

Is there a way of wip'ng out the overview (buffindexed) files, and
history, rebuildling the history so that the system can go live and then
just starting up a process to rebuild the overviews?

Systems Administrator @ hub.org 
primary: scrappy at hub.org           secondary: scrappy@{freebsd|postgresql}.org 



More information about the inn-workers mailing list