HP NCU

[ldoodle]

Hiya,

Got a wery weird problem this morning in that 2 or my DL380 G5 servers, who both have 2x HP NC364T adapters in them, with ports teamed across the adapters, are switching between connected and disconnected on the teams I have setup. Luckily it's only 1 port of each team so nothing is dropping out completely, but what is strange is when both are working, ping stops working, and when 1 drops off, ping starts working.

This is on Windows Server 2008 R2 x64. Literally every 10 seconds or so they drop off and them come back on. Sometimes they go to 'Degraded (Rx path)'. I am using Automatic mode for team, which as i'm using 2 switches as well defaults to TLB.

[Trygve Henriksen]

Are you by any chance using IP V6 on the network cards?
Or have it listed as one of the protocols installed?
(It is instaleld by default... )

If IP V6 is installed, but not used, I've experienced that PING may act 'rather weird'.

What kind of Switches are your servers connected to, and what is the ports set to on them?

Have you checked if the Switches have registered any errors on the ports?

[ldoodle]

Thanks for the reply.

Our switches are NETGEAR GS724TS switches. I can see from the logs that the ports in questions are in a continuous loop of:

Down
Up
Forwarding

and all this looks like it's 30 seconds or so intervals.

I have configured the ports on my switches to match the settings of the team (ie auto neg. etc), not of the individual ports making up the team.

[Trygve Henriksen]

I've seen behaviour like this before...

Basically, the servers and the switches can't agree on transmission speeds.
(Sometimes it can be caused by poor quality wire, power lines close to the cabling, gremlins, or just trying to get two Cisco boxes to talk to each other... )

Try using manual settings on the Switch ports, with whatever speed/duplex you believe will be best for the servers. (Can't be bothered to search out the NIC to find specs... )
Do NOT adjust the settings on the servers, at least not unless you have an ILO-card or easy access to the console...
(That's an error I won't make again... )

[ldoodle]

Thanks I will try that.

Funny thing is though this config has been working happily for the last 2 months.

Came in on Monday this week and started seeing this behaviour on both servers (which are part of a Hyper-V cluster by the way).

[ldoodle]

Oh and isn't it better to manually set speeds at both ends of the chain (ie the servers and the switches) for best performance?

[Trygve Henriksen]

No...

First of all, now you'll be reduced to ONE 'autodetect' routine working on the settings, and that doesn't take much resources at all. (The problem is the autodectects 'clashing' with each other)

Second of all, if you get the settings wrong on the Switch, you can just telnet to it, or use whatever webform is included, but if you mess up on the server, you either need to use an ILO-card or go to the physical console to fix it....

[ldoodle]

Ah OK.

I forgot to say earlier that I don't use IPv6 and the protocol in unticked in the NICs properties (of the team).

[ldoodle]

OK changed the port settings to manual on both switches and it's still doing the same thing!?

Any more ideas?

[Trygve Henriksen]

It may be time to look for 'changes' in the systems.

Did the servers download and install a new M$ 'HotFail'?

Did anyone install new Firmware?

Did someone plug a coffee maker into the UPS?

A new Fluorescent tube lamp?