HP server and VLAN

[Nick M_1]

I have a test environment which I built for the purpose of testing Oracle performance on NAS disk. The Oracle testing was OK, but I had a problem with the HP server because it crashed the VLAN.

The server is a L2000 and the GBe network card is an old A4926A PCI 1000Base-SX Adapter. I suspect the network card and it's drivers are the cause of the problem.

If I take the HP server down or reboot the HP server - all the other servers on the VLAN crash. It seems all the other servers rely on some kind of a keepalive from the HP server. The VLAN connectivity is administered from the cisco switch.

I hope to do some further tests with another GBe card, but before I do this could you answer a couple of questions?

1) Is this a known problem with HP GBe cards on VLAN with cisco switches? Has anyone else had this kind of problem with other models of GBe card in a similar configuration (HP, cisco and vlan)? Are HP servers generally reliable in cisco vlan configurations?

2) My background is in server / database administration (I have a network administrator who does switch admin). Has any one seen this type of VLAN crash problem before? Can you suggest a likely cause of the VLAN crash when the HP server is taken off line? (e.g. network setting, type of polling, etc)

Any help would be much appreciated

Nick

[Marcel Boogert_1]

Nick,

The only thing I can think of is a broken network interface. You could try and replace the network adapter.

Are you using the HP server as a router? If so, that could be your problem.

Can you specify some error messages from the other servers?

MB.

[Nick M_1]

Marcel

These console messages are all I have immediately available (it occurred once before in June).

adausi03 console login:
Jun 25 11:57:31 adausi03 su: 'su root' succeeded for griffirt on /dev/pts/2
Jun 25 15:07:36 adausi03 in.mpathd[125]: All Interfaces in group group_2 have failed
Jun 25 15:08:20 adausi03 nfs: NFS server filer04g not responding still trying
Jun 25 15:08:51 adausi03 nfs: NFS server filer98g not responding still trying
Jun 25 15:12:46 adausi03 in.mpathd[125]: Successfully failed back to NIC ce0
Jun 25 15:12:46 adausi03 in.mpathd[125]: NIC repair detected on ce0 of group group_2
Jun 25 15:12:46 adausi03 in.mpathd[125]: At least 1 interface (ce0) of group group_2 has repaired
Jun 25 15:12:47 adausi03 in.mpathd[125]: Successfully failed back to NIC ce2
Jun 25 15:12:47 adausi03 in.mpathd[125]: NIC repair detected on ce2 of group group_2
Jun 25 15:12:52 adausi03 nfs: NFS server filer98g ok
Jun 25 15:12:54 adausi03 nfs: NFS server filer04g ok

I will try and get some more information

Nick

[Bill Costigan]

Nick,

What do you mean "crashed" did the systems themselves go down, hang or did an application abort?

Your log included messages about NICs in a group, Are you using Auto Port Aggregation, the product that allows multiple NICs to appear as a single NIC or as backups for each other?

Your log also mentioned NFS, are there critical NFS mounts that would cause an application to abort.

I guess the best information would come from the OLDsyslog file from the systems that crashed. That might show what cuased the systems to decide to crash.

[rick jones]

While there may indeed be bugs with vlan support (you _are_ up on all the latest vlan and gelan patches yes?), it also sounds like those other systems have a bug - simply bringing another system down should not in and of itself cause different systems to crash.

[Gerhard Roets]

Hi Nick

Did you check you switch firmware if it is up to date ?

From those logs .... i see there is some nic failovers. Does this imply that you have two nics onto that network ? It might be that the second nic does not have vlan support enabled.

This is just a thought though.

Regards
Gerhard

[Steven E. Protter]

Clearly a networking problem. It may be the NIC, but it may be the software.

Include getting your system up to the latest quarterly update.

Make sure the switch port settings are not set to manual on the switch. That will play havoc with the situation.

If however, the vlan can crash with the hp sever not plugged into the lan, the conlusion is that either the hp server is not the problem or there are muliple problems.

many hp nic cards start to have increasingly long ping times when they are getting ready to die.

SEP

[rick jones]

Ah, Nick - in.mpathd, at least when combined with interfaces named ceN, suggest that the log was from a _Solaris_ rather than an HP-UX system. I do not recall HP-UX having an in.mpathd and the GbE NICs on UX are lanN, not ceN (perhaps a third-party NIC might have a ceN interface name but it is rather unlikely)

Are you attempting to combine APA and VLAN on HP-UX?

[Nick M_1]

Everyone

Firstly, thanks for all your replies to my questions.

I feel encouraged by your replies, they infer that you wouldn't expect a HP, VLAN and Cisco configuration to crash a VLAN. I have a SAN rather than NAS background. I gather Solaris is more established on VLAN than HP, but it sounds although HP is still mature enough on VLAN to be deployed for a production system.

My GBe network card is an old A4926A PCI 1000Base-SX Adapter and it's too old to support APA. Although, from your replies it seems the Solaris hosts on the VLAN are using IPMP.

I have requested some more information on diagnostics and configuration settings, so this is an interim reply. I am not in a situation to do frequent tests on this problem, because it is intrusive & it crashes the entire vlan.

I will let you know when I have more news.

Thanks again

Nick

[Nick M_1]

All

I am going to have to park this for a while because I can't do a test yet. When I take my server off line it crashes all the other servers on the vlan & I have got to wait for a slot that will be less disruptive to other users.

I have awarded some points on the answers I received thus far. I will update the thread when I get more news

Thanks

Nick