Failed LAN card detection.

[Aharon Chernin]

How does Service Guard detect failed lan cards? I see there is a setting for NETWORK_POLLING_INTERVAL.. But, no where I can find an explenation of what it means by network polling. What is SG actually doing when this preconfigured time comes around?

Reason I ask is becuase yesterday I had a failed network card. Well, something failed, becuase the card was not responding. SG did not fail the card over to the secondary card for that lan. So, I pulled the cable out of the bad card by hand, just to humor myself. And then SG failed it over to the secondary lan card..

Something really wierd either happened, or broke in that card. Though, the point is, the machine could not see any devices on that lan. And couldnt even ping the default gateway. Yet, SG still thought the card was good.

So, what the heck does SG do to detect failed lan cards? It had worked fine in the past for me.

Do you think a product like HP Auto Port Aggregation would help in a situation like this?

[Christopher McCray_1]

Hello,

Check out this document:

http://support1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000058670465

and

http://support1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000059363730

Use the keywords "uxsg" or "umcsg" when you search the technical knowledge base.

Good luck

Chris

[Sajid_1]

hello,

Read this document and check the "Understanding MC/SG hardware and software configurations". That describes it:

http://www.docs.hp.com/hpux/onlinedocs/B3936-90065/B3936-90065.html

You can do a search on NETWORK_POLLING_INTERVAL too

[melvyn burnard]

well the question here would be WHAT failed?
if ip failed, but there was still traffic seen on th eport, this would not be seen as a failure.
The network checking is done below ip address, at dlpi level. as long as this gets through, sg sees the network as up.
pulling the cable obviously prevents this, hence the failover worked.

[Aharon Chernin]

Can you explain how IP can fail for your primary card but still work on your secondary card after failover?

I am trying to track down what actually broke myself.

[Stephen Doud]

Hello Aharon,

As Melvyn stated, SG performs the periodic LAN checks at the link-level (dlpi - layer 2 of the OSI model). Hence, an "ifconfig lanX down" can "break" IP (OSI layer 3) traffic, but layer 2 is still happy. This is why testing an ifconfig lanX "down" on a LAN NIC doesn't produce a lan failover as might be expected.

When SG detects a layer 2 outage, it logs it in the syslog.log. The following is an example of a lan failover, followed (after a number of NETWORK_POLLING_INTERVAL cycles) by a LAN recovery:

Feb 27 12:45:16 Node_1 cmcld: lan6 failed
Feb 27 12:45:16 Node_1 cmcld: Subnet 172.28.171.0 switched from lan6 to lan7
Feb 27 12:45:16 Node_1 cmcld: lan6 switched to lan7
Feb 27 13:00:25 Node_1 cmcld: lan6 recovered
Feb 27 13:00:25 Node_1 cmcld: Subnet 172.28.171.0 switched from lan7 to lan6
Feb 27 13:00:25 Node_1 cmcld: lan7 switched to lan6

-s.