From: Rusdi (rusdi@komputeralif.com)
Date: Mon Oct 07 2002 - 23:25:04 EDT
Dear Members...!
I have received 2 replies to my problem.
Thank you very much to:
1. John Galt
2. Bluejay Adamets
I had resolved my problem, what I did, I just
Reset my Decnis routers ( 2 units Decnis, as router and FDDI back bone).
Decnis has an ARP proxy feature to forward packets to different subnet.
I still need to do something on Decnis, so I don't need to reset my Decnis
anymore if cluster service moves to other server.
At lease, now I am sure that the problem is not on cluster's setting.
Regards
Rusdi
--------------------------
John Galt,
Your problem lies with the router located on the subnet with the servers.
This is probably your default gateway.
The arp aging time on the router must be lowered. The mac entry for the
cluster service is registering with the router as the first server's mac.
When the service is moved the second server, the router still uses the mac
address of the first server and the connection from outside your subnet
will fail. Eventually, all entries will time-out, with cisco routers, I
believe the default arp-age is about 4 hours or so. This is way to long
for high availability services.
This is a common problem. You will need to contact the administrator of
the gateway/router on your network and ask them to reduce the arp aging
time on the router.
----------------------------
Bluejay Adametz,
It sounds like there's a router someplace (your WAN router, probably) that's
caching the ARP entry for your service address. When this happens, the
router will just assume the IP address is at the same MAC address it used to
be at, and send the packets there.
We had a similar problem with our Cisco routers, and came up with two
solutions: clear the ARP cache on the routers after a service move, and
reduce the ARP cache timeout value so the routers would re-ARP sooner.
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:48:55 EDT