Big problem, or little problem

N niels at
Thu Sep 9 00:40:29 UTC 1999


>> dbz.c(set): can't tag value 0x1000016 tagboth = 0xff0000
>> Should I panic, or just get a little nervous?  This is compiled with
>> --enable-tagged-hash because the machine's only got 128Mb of physical memory.
> Get a little nervous, but not too much.  :-)  What this means is that
> your history file has gotten a bit bigger than what the .dir/.pag
> files were set for.  Is this a new server, just starting up and
> building up its history file? If so, this sort of thing is fairly
> common.  You do want to run expire soon, though, to rebuild the
> history.dir/.pag to fit the new, bigger history.  Not only is
> overflowing the .pag file inefficient, it does sometimes apparently
> lead to duplicates getting in the history file (I haven't been able to
> track down exactly how that happens), so you probably don't want to
> leave it in that state too long.

What if expire fails with the following error:

Error log:
can't tag value 0x2d521ae tagboth = 0xfe000000

Expire messages:
expire begin Wed Sep  8 23:40:46 MET DST 1999: (-v1 -z/news/log/expire.rm)
[lots of Group not matched (removed?) messages for nuked groups]
    Can't read line 3522193, No such file or directory
    Article lines processed  2508898
    Articles retained           6054
    Entries expired          2502844
    Files unlinked           2874149
    Old entries dropped      1013294
    Old entries retained           0
expire end Wed Sep  8 23:44:46 MET DST 1999

news# head -3522200 history | tail
<rs0fmu7imkf83 at>	935347356~-~935345886
<cancel.8e28.10f9c.28 at default>	935347357~-~935346481	control.cancel/49575656
<8e28.10f9c.28 at default>	935347357~-~935346481	
<7pp3mb$ftt$22 at>	935347357~-~935334411
<pornsexy at>	935347357~-~935280394	
<aa882556.46399858 at news>	935347357~-~935335234	
<7ppga1$c28$7 at>	935347357~-~935339073	telekabel.sprinter.binaries/290260
<aa842552.46398926 at news>	935347358~-~935335233	
<aa772545.46395210 at news>	935347358~-~935335229	
<37E4BC0B at>	935347358~-~935339391	

Ever since this started happening article acceptance is in the low
100.000's per day leading to woefully incomplete threads (let alone
binaries - but who cares about those anyway).  Any idea how I can fix
this?  Run makehistory to rebuild the history file and overview from the
(traditional INN 1.7.2, by the way) spool?


	-- Niels.

More information about the inn-workers mailing list