mbl900@csc2.anu.edu.au (Mathew BM LIM) (11/08/90)
We have a "cluster" of 4 Sun servers, a 4/280 and 3 4/65's. All running
SunOS4.1.
I would like to remotely mount /var/spool/news from the 4/280
on one of the 4/65's and run nntpd and C news on this.
At first I thought that this was mearly a matter of using NFS to
remotely mount /var/spool/news on the 4/65, however it turns
out to be not a easy as all that (isnt it always?)
The nntpd makes a call to statfs(2) to find out the number of inodes
free on the spooling area (or so I gather). If that number is less
than MINFILES, it will not allow a transfer. The problem is (quote from
the SunOS manual for statfs(2) ) :
"BUGS
The NFS revision 2 protocol does not permit the number of
free files to be provided to the client; thus, when statfs()
or fstatfs() are done on a file on an NFS file system,
f_files and f_ffree are always -1."
This causes nntpd to never allow transfers, hence I never get any news
fed to me.
Has anyone got a workaround for this (I have nntp V 1.5.10 by the way)?
I don't mind if nntp never checks for MINFREE as I have about 300MBytes
of spooling space to play with so am not worried about running out
of space. I could hack out the code myself but would rather not.
I have tried using RFS instead of NFS to do the remote mount but (SunOS) RFS has
a strange "Feature" in that it only allows 36000 inodes on a RFS
file system (or so our Sun support person tells me), when that number
is exceeded you start getting strange errors like this :
manuel 12:57pm -> ls /var/spool/news
ls: /var/spool/news/misc: Too many open files
ls: /var/spool/news/psu: Too many open files
ls: /var/spool/news/pubnet: Too many open files
ls: /var/spool/news/talk: Too many open files
./ bionet/ general/ mail/ soc/
../ biz/ gnu/ news@ software/
admin/ canb/ groups/ newsites/ test/
alt/ comp/ in.coming/ newusers/ to/
announce/ config/ junk/ out.going/ trial/
anu/ control/ lists/ rec/ unix-pc/
aus/ ddn/ lost+found/ sci/ vmsnet/
manuel 12:59pm ->
So at most I can only have 36000 articles ??
Why not run nntp on the 4/280 you say? Well I would rather not as the 4/280
is quite heavily loaded and I thought I might offload some of
that onto the other servers. True the network traffic still
ultimately goes to the 4/280 but at least the CPU does not
get flogged to death.
Thanks in advance
---
Mathew Lim, Unix Systems Programmer, Computer Services Centre, ANU, Australia.
Telephone : +61 6 249 2750 | ACSnet : Mathew.BM.Lim@anu.oz
Fax : +61 6 247 3425 | Internet : Mathew.BM.Lim@anu.edu.au
--
---
Mathew Lim, Unix Systems Programmer, Computer Services Centre, ANU, Australia.
Telephone : +61 6 249 2750 | ACSnet : Mathew.BM.Lim@anu.oz
Fax : +61 6 247 3425 | Internet : Mathew.BM.Lim@anu.edu.au