UDEV and HAL Failing on Boot Up on a System with over 2000 Disk Paths (Device Multipather)

[Alzhy]

Our RHEL 5.4/5.5 installations use HPDM 4.4.0 or 4.4.1. Our systems are of course StorageWorks Centric - EVAs and XP Storage.

On 2 of our servers which saw an increase of EVA 8400 and XP Vdisks - we noticed a signfiant hit in OS Startup and noticed UDEV failed (or backgrounded) and HAL failed. But otherwise, evrything seems to be fine and dandy.

One server has well over 2000 disk/device paths in over 512 LUNs(Vdisks). It takes a considerable amount of time to boot.

Any clues what issue(s) UDEV and HAL may have in our situation?

[Ivan Ferreira]

The information here looks related, did you alredy check this:

https://bugzilla.redhat.com/show_bug.cgi?id=459969

http://www.novell.com/support/search.do?cmd=displayKC&docType=kc&externalId=7004877&sliceId=1&docTypeID=DT_TID_1_1

[Pieter 't Hart]

a question comes up

does this host really needs all 512 Vdisks?
or is this host seeing all luns from the other hosts because selective presentation of luns to hosts is not configured on the eva's?
or no zoning configured on the san switches?

[Alzhy]

Ivan thanks.. but that's not it. There's a matching bugzilla but I have not had the time to look at it deeply.

Why 512+ (and growing) vdisks in over 2000 disk device paths? Because this server is our Standby database server and it houses around 4 Databases which are stored on an EVA8400. They were configured very much like how our Production ENvironments that are hosted on XP Storage are configured. -- I know a big mistake but I am having a difficult time convincing our yahoos that this was a mistake and is often the casue of thi server hogging IO on the EV84000 (as this databases are using ASM configuration - so imaging even just one single DB acting up and engaging all 120+ Vdisks all at once!)

We've a case with Redhat but we're still waiting on it. The issue (or non-issue) is Udev fails with a "being backgrounded message" and HAL fails too.. but otherwise, all devices ended up present and nicely multipathed.

I recal there is a bigzilla still out there for the ddevice mapper package that error messages are to spew out on multipath startup or creation and the advice was just to "ignore" if the devices are fine anyway -- worrisome IMHO but this actally still occurs with our very recently patched RHEL 5.4/5.5 systems.

[Pieter 't Hart]

>>> but otherwise, all devices ended up present and nicely multipathed.
<<<

looks like a phantom message ...
it takes UDEv too long to start.
the check for UDEV started, times out and produces the message, but eventually UDEV ends up OK.

[Ivan Ferreira]

Did you try this also?:

http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5/html/DM_Multipath/many_luns.html

[Alzhy]

Yes. We're toying with the ideabut our issue is a boot time issue.

We're following the following though:

https://bugzilla.redhat.com/show_bug.cgi?id=480319

and

https://bugzilla.redhat.com/show_bug.cgi?id=525353