Help Order of NICs in routing table, associated disconnects

[Jupitus]

Folks... this came up at work and I was wondering if any of you techie brainboxes have come across it before. We had a Linux machine with 4 nics installed losing connectivity exactly every four hours.

We found that eth4 was 'higher up' the routing table and conflicting with eth2, and when eth4 was disabled the problem went away.

Question is... why exactly every 4 hours??

 

[TdC]

how can two different nics conflict in the routing table? what do you mean by that?

The order in the table doesn't matter, as routes with a longer prefix take priority over more general ones. The most specific route always wins.

 

[Jupitus]

Well the routing tables had this:

159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth4
159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth2

and

159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth2
159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth4

the first being a server with the problem, the 2nd server did not... why on earth they would have the same entries for 2 nics in the first place I really don't know !!

 

[Trem]

Its something to do with pastry and carpet. Possibly a connection to midgets is also likely.

 

[old.Osy]

Well the routing tables had this:

159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth4
159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth2

and

159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth2
159.220.108.0 0.0.0.0 255.255.255.128 U 0 0 0 eth4

the first being a server with the problem, the 2nd server did not... why on earth they would have the same entries for 2 nics in the first place I really don't know !!

Amateur bridging attempt?

 

[Jupitus]

I've just heard that the idea is for eth4 to be a backup NIC to eth2, hence the same routing...

 

[Jupitus]

Oh... anyway, the main point is why the 4 hour interval?

 

[TdC]

I've just heard that the idea is for eth4 to be a backup NIC to eth2, hence the same routing...

um...bonding interface? fire your server admin.

 

[caLLous]

See if there's a cron job scheduled every 4 hours that could be triggering something?

 

[Jupitus]

Nope, no crons...

 

[Moriath]

Nope, no crons...

Just an old crone.

 

[TdC]

well Jupilar old chum, if you have both those NIC's active, and they're in the same subnet, then I am kind of inclined to think that all kinds of bad networky things could happen with an identical route on both interfaces. if, as you mention, the goal is to have a redundent or standby interface, I'd certainly read up on the linux bonding interface.

ofc you could mean something else entirely. I have a fairly decent working knowlege of IP based networks, but anything exotic-ish is well beyond my expertise.

 

[Jupitus]

No worries Teedles, as I said the guys found a resolution, I was really just wondering if anyone had deep enough experience here to perhaps know why it would have an impact at 4 hour intervals

 

[TdC]

I must have missed that bit in your post chum. Sorry for banging on about it

 

[Jupitus]

I must have missed that bit in your post chum. Sorry for banging on about it

No worries, my little cock-obsessed study-wudy