fullfeed related questions / cycbuff limitations
hscholz at raisdorf.net
Wed Sep 10 10:24:12 UTC 2003
I've set up a 8CPU 500MHz XEON box which should receive a full
feed including binaries and store them on a NFS storage using
innd seems to be limited by the fact that the incoming feeds have
to be handled by one process. The backlogs of the remote
servers take >12h to be processes and might increase on one side
when innd is busy handling the other one.
The db/ directory is on a local raid, spool is on NFS with Gbit.
Disabling tradspool for text groups does not increase performance
since the network load is only ~10MBytes/s.
Is a 500Mhz Xeon (1MB Cache) enough for such a fullfeed?
Would a 700MHz Xeon (2MB) handle it?
Btw: Are there any reasons for the current cnfs limitations (12 buffers
Is there a way to group metagroups into larger metagroups?
I have ~15TB storage in 800GB chunks (limitation of the storage).
How should I do the layout for these?
Hendrik Scholz - <hscholz at raisdorf.net> - http://raisdorf.net/
drag me, drop me - treat me like an object
More information about the inn-workers