lan5 failed

[Davor Bira?]

I have ServiceGuard OPS Edition Bundle ver. A.11.14. My problem is standby network interface: lan5.
It repeats this message in syslog.log:
Mar 29 08:15:43 XXX cmcld: lan5 failed
Mar 29 08:15:49 XXX cmcld: lan5 recovered
Mar 29 08:43:38 XXX cmcld: lan5 failed
Mar 29 08:43:44 XXX cmcld: lan5 recovered
Mar 29 08:48:42 XXX cmcld: lan5 failed
Mar 29 08:48:48 XXX cmcld: lan5 recovered

Is this something to worry about?

[Duncan Edmonstone]

Absoluteley - you have an intermittent network issue - have your network staff take a look at the switch that lan5 is attached to...

HTH

Duncan

[Isralyn Manalac_1 · Accepted solution]

It is indeed something that needs your attention. It is best to liaise with the network team to check the health of the switch with which this cable is connected to. Another log file worth checking is the nettl.LOG000. Do a ps -fea|grep netfmt and you'll find that file.

Regards,

Isralyn

[melvyn burnard]

Well yes, as this is telling you there appears to be an intermittent lan problem that could render a service, a package, or worse a node as failed.

get your networking folks to sort out this issue as soon as possible.

[doug mielke]

There is a clue here in that each of the 3 events lasts 6 seconds.

This looks like some sort of 'discovery' like network flood. Too short to be a legitamate network overload, consistant enough to indicate an intentional network task.

If you get no diagnosis help from your network folks, you could try asking them to block some of the more common broadcasts from these switch ports and allow only what ServiceGuard requires.

[Stephen Doud]

The lan failure/recovery period is determined by the NETWORK_POLLING_INTERVAL designated in the cluster configuration ASCII file. The default value is 2 seconds.
You can verify what the setting is in your cluster by inspecting the cluster's binary file:
# cmviewconf | grep -i polling
network polling interval: 2.00 (seconds)

Since the recovery is 6 seconds after the failure, 3 check periods occur (on the default setting). The check is performed at the link-level, so transmission capability is being compromised. Best to get a network sniffer on the line and see what's happening.

[Davor Bira?]

At the end it was problem with switch, so thanks to all for replays.

[Davor Bira?]

Thank again