Is buffindexed overview corruption problem fixed in -STABLE?

[bill davidsen]

In article <Pine.LNX.4.33.0103281344190.16782-100000 at malvix.hist.no>,
Mads E Eilertsen  <mee at hist.no> wrote:
| 
| On Fri, 29 Sep 2000, Katsuhiro Kondou wrote:
| 
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: ovgroupmmap ovbuff is null(ovindex is 13112, ovblock is 842019127
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: could not open overview for 'control.cancel'
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: ovgroupmmap ovbuff is null(ovindex is 2353, ovblock is 875575352
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: could not open overview for 'junk'
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: ovgroupmmap ovbuff is null(ovindex is 13616, ovblock is 825242424
| > Jul 12 02:33:31 server expireover[5109]: buffindexed: could not open overview for '1.beginners.sex'
| >
| > After creating buffers as regular files, those are gone.  I'm not
| > sure but regional lock on block device may not work.  Are still
| > someone there who get those messages on regular files?
| 
| Yes,
| with inn-2.3.1 on FreeBSD-4.1.1 and Compaq Tru64 UNIX 5.1,
| with buffers as regular files.
| 
| The newsgroups seem to be the same from day to day, with a few
| new groups added to the error list each day.
| 
| Is this something to worry about?
| 
| BTW, I have 40 buffers of 100MB each.  Would I gain something by using
| fewer and larger buffers?

As far as worry about, I don't think so. I almost remember what this
shows. Pick one or more, a group removed from active, or a group in
active which does not go to overchan. If you have something like
  controlchan:*,!control.*,control.newgroups,!junk
or similar, you (a) may cause this problem, and (b) are saving a lot ovf
overhead. Sounds like a good tradeoff to me.


-- 
bill davidsen <davidsen at tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.