On 7/7/12 1:24 AM, "Jared Mauch" <jared@puck.nether.net> wrote:
Die proxy arp die. (and that's not German).
I've had a job or consulting gig or two that has inadvertently had this as the hidden glue making things work.
(wha, you can't route that subnet out an Ethernet interface without a next hop? It's always worked....)
I fight with sysadmins to this day about the concept of a broadcast domain and subnet... If I hear another case of someone saying that switch is the "80" subnet when there are 3 co-existing /24s in that domain I may go crazy....
I've cleaned up a lot of poor host and network management and it's amazing how much a difference the hardware operates without the hacks.
Jared Mauch
On Jul 6, 2012, at 8:51 PM, Ben Aitchison <ben@meh.net.nz> wrote:
Routing loops, incorrect subnet masks. (like when people stick a /24 netmask on a /27 then can't reach another adjacent /27)
We had a pair of diversely located systems operate for about 18 months with misconfigured gateway addresses. Proxy ARP kept everything on an even keel until one of the systems failed and the traffic routed to the remaining system. I arrived on the call in time to hear the sys admins saying that they had exceeded the maximum number of ARP entries and were going to expand the table :(