topic Re: DL580 & NC7170 link problem (RH3) in Operating System - Linux

DL580 & NC7170 link problem (RH3)

Timo J — Thu, 18 Aug 2005 04:27:31 GMT

I'm having continuous link problems with those NC7170 nics. Configuration is as follows:

- two-node ServiceGuard cluster
- bond0 (eth0 & eth4)
- bond1 (eth1 & eth5)
- and of course, both nics belongins to same bond are connected to separate switches.

If you look the attached file you can see that links go down very often, and come up on same second. Am I having faulty hardware or what?

Should I increase the value of 'miimon' on bond configuration to get rid of those messages? (miimon is now 100ms). That doesn't sound very reasonable solution.

Hardware configuration is identical on both nodes, but the problem exist mainly in first node.

Re: DL580 & NC7170 link problem (RH3)

Timo J — Thu, 18 Aug 2005 05:04:38 GMT

Ok, technical knowledge base helped once again: http://www5.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000078411083

Re: DL580 & NC7170 link problem (RH3)

Tvs — Thu, 18 Aug 2005 06:29:08 GMT

hai

try to decrease the speed of the nic card by mii-tool

Re: DL580 & NC7170 link problem (RH3)

Durgaprasad Ala — Thu, 18 Aug 2005 14:07:33 GMT

We have almost 300 servers running bonding
and we get these errors often and it is not going affect the performance.

We never face any connection loss or performance degradation.

Also if you are using 2 switches,Ask your Network engg to trunk(ISL) the switches.

Re: DL580 & NC7170 link problem (RH3)

Timo J — Tue, 23 Aug 2005 01:31:12 GMT

Looks like I had to increase the value of miimon. As you can see from attached file, the bond0 goes down, and ServiceGuard notices that bond0 is down and initiates a failover (back to same node, but still a failover.) I'm not using the option use_carrier.