Question about "<news>/db/history" file size

[Petr Novopashenniy]

On Thu, 23 Apr 2009, Russ Allbery wrote:

RA> Petr Novopashenniy <pety at rusnet.ru> writes:
RA> 
RA> > But expire process uses ~1 GBytes of memory.
RA> >
RA> > When history file was ~4,5 GBytes (two weeks ago), expire process uses
RA> > more than 1,2 GBytes of memory. Yes, looks like "expire process size"
RA> > = ~"history.index size".
RA> 
RA> Yeah, innd has more locality to its writes, but expire writes all over
RA> the place.
RA> 
RA> > Maybe it is possible not do full memory map (at one time) for history file 
RA> > index (for expire process)? 
RA> 
RA> It'll be dog-slow.

Maybe there are other algorithmic solutions? Since swapping is not working 
at all :(.


RA> 
RA> > And, also, it is possible show number of articles in cnfs buffers (in 
RA> > "cnfsstat -a" output)?
RA> >
RA> > Something like this:
RA> > Buffer news1, size:  47.7 GBytes, position:  28.3 GBytes  10.59 cycles
RA> > Articles count: xxxx?
RA> 
RA> I believe the CNFS metadata doesn't actually store that information.
RA> 
RA> > And, if expire process not shutdown correctly, innd remains in "reserved" 
RA> > mode. How I can "ureserve" innd without restart?
RA> 
RA> ctlinnd go 'reason matching the reserve reason'

Oh! I not used 'reason matching the reserve reason'! 
It's my mistake, thanks!


RA> -- 
RA> Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

--pety