At 10:36 AM 6/19/01 -0500, you wrote:
>Thanks Chris,
>
> That is how I have my disks set up.. everything except some spools are
>local, some spools are NFS mounted NAS. The bandwidth available for NAS
>should be far in excess of disk performance. The network links are 100BaseT
>full duplex directly connected to a very lightly loaded Cisco 6509 switch
>(the 1% load light has yet to illuminate) which is directly connected to my
>DS3. I just received a fiber gigabyte interface this morning and may put
>that in the news server to see if it changes things.
>
>The speeds are somewhat better since I put in the new build and implemented
>the config changes you suggested in another thread. They now sometimes range
>into the low hundreds of K/s, but still not what I would expect on a system
>with comparatively low bandwidth utilization (my DS3 normally runs at about
>2-10M/s).
Ok, it looks very much like your problem is not dnews, here your stats shows:
Busy 2.2% for the last 4833 seconds
Which is a sure sign that dnews isn't in trouble and you don't have a disk
IO problem.
So this leaves three likely problems
1 Network problems
2 Sending end at fault.
3 Protocol problems
I'm 99% sure '2' is the problem but it's probably worth doing some tests to
be sure.
However, first you should look into this, see the 'rejects' figure is huge.
Posts Rejects Dup Total Articles/S Kbytes/s K Site
1195 14069 20 15284 3.17 223.29 1075827k 63.21%
newsfeed-central.sprintlink.net
That probably means your active file doesn't match, anyway examine dnews.in
and check why you are rejecting
so much of the feed.
There are 2 ways to prove your incoming feed is the problem,
1) Get a trial feed from someone else
2) Send in a test feed using the test program described below. all
this does is shows you that
your server can take news faster, and you can run it from outside your own
network to test your
widearea link.
ChrisP.
ftp://ftp.netwinsite.com/pub/dnews/feedtest
You can use this utility with a command like this:
tellnews newgroup test.feedtest y
In dnews.conf increase 'too_old' to 2000, temporarily
tellnews reload
./feedtest -group test.feedtest -to site.name:port -size 40000 -items 500
It will print out figures showing the speed it could send news into your
system. By
testing your server on the server itself and then again on another server
in the same LAN,
and again outside your LAN on the WAN, you can establish if the server or
your network or
the wide area network is the problem.
WARNING: Always be careful to send this test feed to a local TEST news
group like test.feedtest !!!!.
.
>Thanks again for all the great help and support!!!!!
>
>Kevin
>
>"NetWin Support Auckland" <[EMAIL PROTECTED]> wrote in message
>5.1.0.14.2.20010618093417.01ae2430@ball">news:5.1.0.14.2.20010618093417.01ae2430@ball...
> > At 08:03 PM 6/15/01 -0500, you wrote:
> > >I am interested in the answers to this myself. I am using NAS storage
>with
> > >my DNews server. It certainly appears to be stable, though I am not
>seeing
> > >incoming feeds over 100K/S. I have been working on this as a seperate
> > >issue, but your comments make me consider the possibility of the NAS as a
> > >botleneck.
> > >
> > >My server is a dual P800 running Linux with a 100BaseT connection to a
>170G
> > >Quantum SNAPServer for NAS.
> >
> > A NAS solution can work, and it can be a bottleneck. In general I would
> > recommend
> > organizing it like this if you don't actually need to share the data.
> > Workarea (local)
> > xover (local)
> > history (local)
> > spool1..n (NAS)
> > That will work best as all the high speed low latency transfers will be
> > local and the
> > bulk storage is remote.
> >
> > But it can infact work ok with everything remote as long as your system is
> > properly
> > configured. For example you probably want the network link working in
>full
> > duplex mode
> > if possible. And the nfs mount point should be configured with write
> > confirmation disabled
> > if it's an option (it is on NT I can't recall with NFS)
> >
> > Anyway, by all means send me the output of:
> > tellnews stats
> > tellnews status
> > tellnews stats_in
> > And I can give you a rough idea if the disk IO is your problem, in brief
> > you look for
> > a busy figure above 90% and a big difference between cpu and elapsed
> > columns in the
> > stats output.
> >
> > But really 100K per second is very slow it's unlikely to be caused by your
> > system but more
> > likely the sending end or the wide area network speed.
> >
> > ChrisP.
> >
> >
> >
>
>
>
>
>
>
-------------------------------------------------------------
To unsubscribe send to [EMAIL PROTECTED] a message
containing a single line, unsubscribe
Problems with:
Incoming feed problems: http://netwinsite.com/dnews/feedsin.htm
Access rule problems: http://netwinsite.com/dnews/access.htm
General FAQ: http://netwinsite.com/dnews/faq.htm
