topic Re: Problem with server in Operating System - HP-UX

Problem with server

80skeys — Fri, 13 Feb 2009 22:41:59 GMT

We've got a strange problem with our server. For the past two days we've been working with both HP and Sybase tech support, but they have been unable to solve it. I'll describe the problem as briefly as possible:

Server: HP-9000/800 B.11.11
Symptom: "load" from tape of sybase databases larger than 20GB extremely slow, load eventually hangs around 100GB, progressively slower, database fails to go online. For first 2-3 GB of data, load goes fast, then immediately slows to a crawl.
Error logs: none

Through standard troubleshooting, we have ruled out the following: tapes, tape drives, network.

Steps taken:
(1) used "glance" to check hardware.
(2) reinstalled sybase.
(3) rebooted server
(4) looked through system logs
(5) sent performance tool results to HP

Nothing appears out of the ordinary.

Any ideas?

Mike

Re: Problem with server

Steven Schweda — Fri, 13 Feb 2009 23:19:56 GMT

> [...] "load" [...]

"load" how? Actual command(s)?

> [...] tape [...]

What kind of tape?

> Through standard troubleshooting, we have
> ruled out the following: tapes, tape
> drives, network.

Procedure? Evidence?

> [...] as briefly as possible:

Too.

Re: Problem with server

Avinash20 — Sat, 14 Feb 2009 07:26:13 GMT

Load eventually hangs around 100GB.

Do a TOC and provide the crash to HP Solution center.

MP> CM > TC

Crash dump will be having much information and may lead to some conclusion.

Re: Problem with server

80skeys — Wed, 18 Mar 2009 13:25:14 GMT

This has been a tough one, but I have finally identified the issue. I ran the Offline Diagnostics Environment (ODE) which showed a slew of errors on ikediag2 which tests the IKE IO controller subsystem on the server. As this is a hardware problem, the next (and only) step is to ask HP for resolution.

Re: Problem with server

Torsten. — Wed, 18 Mar 2009 16:06:41 GMT

It was a bit hard to find this thread again, you opened so many about this problem ...

As said earlier (in another of your threads?), >most< hardware problems should be detected during the selftest, but not all.

Once you managed to capture the test output, please post it.

The IKE chips (seen as SBA - system bus adapter - in ioscan) connect the CPUs to the IO, one chip is responsible to connect each PCI backplane.

A problem there can cause problems like yours, IMHO.

Re: Problem with server

80skeys — Wed, 18 Mar 2009 17:03:19 GMT

ODE log output attached.

Re: Problem with server

Torsten. — Wed, 18 Mar 2009 17:16:50 GMT

From a very first look this may be a problem with the card at path 0/5/0, this is slot 1 (left I/O backplane).

What card is in tis slot? Looks like a SCSI card. I would first try to replace this card ...

Re: Problem with server

80skeys — Fri, 03 Apr 2009 16:12:15 GMT

This was solved by replacing the original VA7410 and on the new array setting the QFT to 75.