Clusterware on 11.31

[Oviwan]

Hey folks

I'm installing oracle clusterware 11.1.0.6 on a BL860c 11.31. But the server crashes if I start clusterware, so I opended a SR. they told me I have to run the cluvfy script ($ORA_CRS_HOME/bin/*cluvfy stage -pre crsinst -n -verbose). and this is the "interesting" output:
cluvfy output:
---------------------
Check: Package existence for "ACXX:C.06.12"
Node Name Status Comment
------------------------------ ------------------------------ ----------------
ex10438 failed
ex10437 failed
Result: Package existence check failed for "ACXX:C.06.12".

Check: Package existence for "C-ANSI-C:C.06.12"
Node Name Status Comment
------------------------------ ------------------------------ ---------------
-
ex10438 failed
ex10437 failed
Result: Package existence check failed for "C-ANSI-C:C.06.12".

Check: Package existence for "FORTRAN90:B.11.31.01"
Node Name Status Comment
------------------------------ ------------------------------ ----------------
ex10438 failed
ex10437 failed
Result: Package existence check failed for "FORTRAN90:B.11.31.01".

Check: Package existence for "make:3.81"
Node Name Status Comment
------------------------------ ------------------------------ ----------------
ex10438 failed
ex10437 failed

so on hpux I have to pay for these software, right?
Ok, the make package I found on http://hpux.connect.org.uk/hppd/hpux/Gnu/make-3.81/ but what is with the others?

I have already an other cluster on hpux with clusterware and these packages are not installed there... so this should not be the cause?

Thanks

Regards

[Fabio Ettore]

Hi,

it sounds like the software requirements for the product are not satisfied.

Installation Guide of Oracle Clusterware 11g for HP-UX:

http://download.oracle.com/docs/cd/B28359_01/install.111/b28259.pdf

Software Requirements List for HP-UX Itanium Platforms


It asks for Fortran and aC++ Developer's Bundle, are they onto the system?

Best regards,
Fabio

[Oviwan]

hey

thanks for reply.

no they are not installed:
swlist | grep -i fort

where can I download them? are they for free?

Regards

[Fabio Ettore · Accepted solution]

Hi,

I suggest doing always

swlist -l product

to search for more detailed installed products, just swlist doesn't report products contained in bundles.

However it sounds you don't have aC++ and Fortran. Now I reply to your first question:

---> so on hpux I have to pay for these software, right?

Yes, it's right.
You can get them by software.hp.com. Just insert in search line aC++ and fortran.

HTH.

Best regards,
Fabio

[Dennis Handly]

>where can I download them? are they for free?

You have to pay for these compilers. But if you aren't going to do database programming in these languages, Pro*C, you should just ignore these errors. Oracle is notorious for mentioning languages patches when you don't have the product installed.

[Hein van den Heuvel]

>> But the server crashes if I start clusterware, so I opended a SR. they told me I have to run the cluvfy script

That SR is an Oracle SAR?

If so, I wouls also excalate to HP.
It does not seem reasonable to me for the (hpuc) server to crash no matter what silly thing the Oracle code tries to do.

Just an opinion,
fwiw,
Hein.

[Oviwan]

hey

yes an oracle tar or sr...

I found this on hp:
http://www.itrc.hp.com/service/cki/docDisplay.do?docLocale=en&docId=emr_na-c01446478-2

my env is also a BL860c with hpux 11.31...

when the server crashes, the last message I see is:
ORACLE CRS (OPROCD) TOC FOR CLUSTERWARE INTEGRITY ...

it also writes a crash dump. can you show me shortly how I can see this dump in a readable format as in DOC: emr_na-c01446478-2?

Regards

Regards

[Fabio Ettore]

Hi,

I'll spend my 2 cents to try to make the problem clearer.

Hein:
"It does not seem reasonable to me for the (hpuc) server to crash no matter what silly thing the Oracle code tries to do."

Oracle CRS holds a possibility to TOC the HP-UX system for Clusterware integrity. That is done by reboot_toc (also seen in the document posted by Owiwan, the document shows a known issue of Oracle CRS 10g when having two init.cssd and it has to referred to Oracle support, anyway it was on 10g and fixed by patchset 10.2.0.4). Owiwan if I see well, you have 11g.

However it's expected a TOC if Oracle CRS manages such condition, and in Owiwan case I see something like that:

"when the server crashes, the last message I see is:
ORACLE CRS (OPROCD) TOC FOR CLUSTERWARE INTEGRITY ... "

I suppose that the analysis of dump will show reboot_toc did the reboot of the system and Oracle should be involved to understand better the reason. Anyway you could consider to open a HP call to get tools for analyzing the dump (something like crashinfo), they will take a look into the dump and (probably) they will suggest opening a call with Oracle.

HTH.

Best regards,
Fabio

[Hein van den Heuvel]

A yes, it did mention TOC (supposedly Transfer-Of-Control, not Table-Of Contents).

Still, it seems wrong to me.
If Oracle gets to crash the box, then it better had been given the explicit authorization to do so (during install? in config?)
Who says oracle is the only game in town.
Admittedly it often is, but the system could easily be running something more important where the integrety of an Oracle (test) instance is less important than the other stuff the box is doing.

fwiw,
Hein.

[Fabio Ettore]

Hi Oviwan,

what was the end of the problem (if ended)?

Best regards,
Fabio