some general questions
paul at atgi.net
Wed Sep 8 20:12:28 UTC 1999
i'm running inn-1999-09-07_03-00/solaris 7. using raw CNFS for both
articles and overview, with a small tradspool for some local/regional
groups. just brought this server online in the last week or so.
yesterday i tried 'moving' one of those regional groups onto tradspool,
that is, i added it to this list of newsgroups in the tradspool
declaration. previously it was being spooled with all the other articles
under the CNFS catchall declaration. however, after restarting, innd would
run away after just a couple of minutes, consuming all CPU, but doing no
work (a truss showed it stuck at 'time()' ). i presume that this has
something to do with articles existing in CNFS but being listed as
belonging in tradspool. I'm not sure how to deal with this without just
doing a dd of /dev/zero to the CNFS buffers, which i'd rather not do. I've
read the man pages for storage.conf, etc, but didn't see anything relevant
to actually changing the declarations of an already-in-place metacycbuff.
The other issue that's come up is that my many CNFS buffers for catchall
are already cycling, after only about four days of spooling. I'd hoped for
a longer retention time. Has anyone used the SEQUENTIAL mode for CNFS
buffers, and would this have the effect I think it might, that is, articles
would tend to stick around longer, since the 'oldest' CNFS buff would tend
to not be written to as frequently? maybe? or perhaps it would be better to
reallocate my cycbuffs, and 'partition' the heirarchies better (currently,
fourteen 1.8G buffers for alt.binaries.pictures*+alt.binaries.multimedia*,
fourteen 1.8G buffers for alt.binaries.sounds*, and fourteen 1.8G buffers
sorry for chattering on so much.
Paul Theodoropoulos Advanced TelCom Group Inc.
Senior Unix Admin Internet Services Division
paul at atgi.net Santa Rosa, California, US
Work: http://www.atgi.net Play: http://www.anastrophe.com
More information about the inn-workers