Olaf,
You should leave probing in -- but test your code more thoroughly and in
multi-vendor configurations. I am also wondering if SLP (Service
Locator Protocol) is supported on most routers today, including linux,
and if there's an alternate method to using ICMP to discover a
gateway/DNS/DHCP server address with SLP. It would be preferable to
using a spamming method. I've also noticed that even when you probe
without the network barfing, LIZARD will get confused and replicate the
same address across several machines (three machines installed with 2.3
all claimed 207.109.151.240 as their address, then promptly crashed my
FTP server).
It's also OK to just let folks input their own IP addresse, but you
shouldn't give up on getting this feature right. MS NT/2000 does NOT
provide this level of capability, in their install (except for DHCP).
The following produtcs all exhibit this problem, and can crash segments
of the internet if an MS NT sever is being used as the Gateway, DNS or
DHCP server on the same Network with Caldera OpenLinux and Open Linux
talks to it during a LIZARD install:
OpenLinux 2.2 Desktop
OpenLinux 2.3 Desktop
OpenLinux eServer 2.3
I tested all three this weekend, and successfully crashed our T1 and a
portion of the internet in Salt Lake City for about 5 minutes for each
test run. I was surprised to see 2.2 crash as well. I also tested the
latest 2.3 on Caldera's Website, and it also will crash internet users
by generating massive spamming packet storms.
In other words, everything Caldera has shipped in the past year or is
shipping now seems to have this problem, and is a toxic combo if
combined with any MS NT/2000 servers on the same network.
This need to get fixed ASAP. The only solution for customers at present
with mixed networks is to unplug the networking cable from the back of
the system during LIZARD install, manually enter the TCPIP info, then
reboot the machine prior to attempting to plug it into a live network.
Jeff
Olaf Kirch wrote:
>
> On Sun, Feb 27, 2000 at 04:42:20PM +0000, Alan Cox wrote:
> > The gateway probe is tricky. You may start routing through one of the millions
> > of machines erroneously shipped by vendors with routing enabled or via the
> > wrong router. There is a router discovery protocol. Its less supported but a
> > lot safer. Listening for OSPF/RIP/RIP2 can also do the detection sometimes.
>
> Hm, ICMP router discovery never worked for me. From the comments in
> Linux's icmp implementation I understand that this is supposed to
> be handled by gated. But from my experience only very few people do,
> and if they do, this rarely happens on internal routers. That's also
> why you rarely observe OSPF or RIP on your department LAN etc.
>
> > At the minimum you need to verify that the icmp time exceeded comes from a
> > host which you cannot directly reach. If you can directly reach the 2nd or
> > 3rd hop that errors then you have found a secondary host misconfigured as
> > router, or a secondary router, and it just told you the correct router to
> > use. Sometimes you'll get a redirect, sometimes you dont.
>
> Good idea. That should also cover the case where you two routers
> playing ping pong with your packets.
>
> Concerning suggestions made by others: netprobe already does much of
> this. For instamce, the initial probe just sits there passively,
> listening to packets going by and trying to figure out DNS servers, the
> local subnet, gateways, and NIS domain. But this is becoming increasingly
> useless as more and more people start to use switching hubs.
>
> Also, DHCP is a nice protocol, lizard does support it, and we do rely
> heavily on it in our test environment, but it's really not being used very
> widely (as just about every other configuration discovery protocol). For
> some mysterious reason, many site admins do seem to prefer to walk around
> and configure their machines manually.
>
> As to the question whether probing the network is acceptable: Prior to
> this DNS probe fuck up, I have had only positive feedback on this
> feature. If customers start complaining netprobe will be removed.
> I'm definitely in the mood to make it less intrusive.
>
> Cheers
> Olaf
> --
> Olaf Kirch | --- o --- Nous sommes du soleil we love when we play
> okir@monad.swb.de | / | \ sol.dhoop.naytheet.ah kin.ir.samse.qurax
> okir@caldera.de +-------------------- Why Not?! -----------------------
> UNIX, n.: Spanish manufacturer of fire extinguishers.
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 21:00:20 EST