[leafnode-list] Can't get leafnode to work with news servers that work with clients.
Matthias Andree
matthias.andree at gmx.de
Fri Jun 10 00:40:21 CEST 2005
Adam Funk <adam00f at ducksburg.com> writes:
> The news version has solved problem 2: the processes now all finish within
> a few minutes. The syslog info and the mail spool, however, indicate
> that my ISP's news servers are disconnecting before fetchnews gets all
> the desired news. I assume that is my ISP's fault! Would it be possible
> for fetchnews to attempt to reconnect instead of aborting when this
> happens?
I'd like to do something like this for leafnode-2.0, but I don't think
I'll change leafnode-1.11 in this regard, this requires a bit of code
and the changes are nontrivial, and requires diligent planning of the
retry behavior - we can't just retry the same server right away without
trying the others first (if more than one is specified), to avoid DOoS,
and we certainly don't want fetchnews to run forever if the upstream is
really wedged, for this might get expensive on modem lines.
This is, I believe, too large a change for the stable leafnode-1.11
branch.
> I've increased the cron fetchnews from once to three times an hour, and it
> seems to be getting a greater proportion of the news now. I may increase
> it again and see how it goes.
Are you filtering ICMP traffic on a firewall, router or via packet
filter rules; particularly on lines with MTU lower than the default MTU
of your network interface (PPP over Ethernet!) or congested lines?
If so, make sure that PMTU discovery (PMTU = Path MTU, MTU = Maximum
Transfer Unit, think "supported packet size") works (there are web sites
on this topic, these should list which ICMP traffic you need to allow).
If you're sure PMTU works (for instance because you're letting all ICMP
traffic pass in either direction), contact blueyonder support.
HTH,
--
Matthias Andree
More information about the leafnode-list
mailing list