cnfs cycbuff problem

[Kjetil Torgrim Homme]

[ab at icsmedia.de]

>   I defined a test cnfs with 50MB. The cnfsstat sayed befor the first cycle:
>   
>   	 Buffer REST02, len: 50.00 Mbytes, used: 48.31 Mbytes (96.6%)   0 cycles
>   	  Newest: 2000-04-11 11:50:38,    0 days,  0:00:01 ago
>   
>   And in the next cycle:
>   
>   	 Buffer REST02, len: 50.00 Mbytes, used: 12.10 Mbytes (24.2%)   1 cycles
>   	  Newest: 2000-04-11 11:53:12,    0 days,  0:00:01 ago
>   
>    
>   After that lines, I think I misunderstand the cycbuff.  Is that
>   correct, that the innd erase all exist messages in a store , if he
>   reached the next cycle ? I think that were really bad.

No, the cnfsstat output is unclear, sorry about that :-/ You're not
the first to be confused by this.

It just means that it is 24% into the second cycle.  CNFS won't
overwrite more articles than necessary.

Perhaps it would be better to say:

 Buffer REST02, len: 50.00 Mbytes, used: 48.31 Mbytes    0.97 cycles
 Buffer REST02, len: 50.00 Mbytes, used: 12.10 Mbytes    1.24 cycles

The "used:" can still be confusing, though.  "at:"?

>   At second, We have a 240G newsbase, with 120x2G container and a
>   800M history.  After a power blackout, I started a "makehistory
>   -buv" , "makehistory -O" , a "ctlinnd renumber" and a normal
>   expire. Normaly I think, that was all to fix the newsbase. But
>   after days we got allready wrong sorted MSGs . I mean , if I read
>   the group comp.security.firewalls , I got MSGs from comp.* .

Overview data, or actual articles?

I use commands like these to rebuild overview (when the overview
partition is full, the overview is corrupted and you need to rebuild)

$PATHBIN/makehistory -I $PATHTMP/newindex -O -n 
$PATHBIN/makehistory -o -r -s `wc -l $PATHDB/history`
sort -t' ' +1 -2 $PATHTMP/newindex > $PATHTMP/newindex.sorted
$PATHBIN/expireindex -a -o -z < $PATHTMP/newindex.sorted


Kjetil T.