Ocfs users - Nov 2004 - Re: TAR 4182518.999: OCFS problems encoutered during support Intel project

Hi Dean,

Thanks for your kind help.

Since I couldn't use vpn in Intel and be a little late after I arrived 
home, I didn't
get chance to modify the TAR on metalink. Tomorrow, I will go CDC and
update it accordingly.

We think all the problems should be caused by OCFS itself. There are two
outstanding problem for now.

First, we will get below error message when we issue "srvconfig -init"
command
to initialize OCR.

oracle.ops.mgmt.rawdevice.RawDeviceException: PRKR-1064 : General 
Exception in OCR
       at 
oracle.ops.mgmt.rawdevice.RawDeviceUtil.<init>(RawDeviceUtil.java:136)
       at 
oracle.ops.mgmt.rawdevice.RawDeviceUtil.main(RawDeviceUtil.java:2071)

I think this is one bug of OCFS for IA64 machine. I will ask Colin to 
confirm
whether there be available fix or patch for this problem for OCFSv1 on IA64.

Second, it will hang at 37% when we issue DBCA to create database. We found
it hangs when it begin to create redo log files. Also need to ask OCFS 
team for
help.

I tried two version of OCFS for IA64. They are 1.0.13-11 and 1.0.11-1. Of
course, we could exclude the possible reason of mismatch between linux 
kerenl
and that needed by OCFS driver. (We first used RHEL3 Update4, kernel
version is 2.4.21-20 while OCFS driver want 2.4.21-4. Then I reinstalled OS
with Update 1, kernel version is 2.4.21-20, we found OCFS driver need 
2.4.21-
4.102 something like that). I think maybe that is too strict for OCFS 
driver to
use.

Please give your suggestion.

Best regards,

Zhihui.

Dean Tan wrote:
> Now that should've some idea of how the problem occur, here's few 
> suggestions.
> - Do you have all the required OS and Oracle s/w version, and patches?
> - Search ITS/WebIV/Metalink on similiar cases
> - Maybe set the owner of the TAR back to the Oracle Support person who 
> was trying to help us earlier, and work with him.
> http://webiv.oraclecorp.com/cgi-bin/webiv/do.pl/Get?WwwID=tar:4182518.999
>
> I've seen no update on this TAR, any problem you're having?
>
> regards,
> dean.
>
> xiaohui he said the following on 11/28/2004 10:39 PM:
>
>> Zhihui wonder if there is some mismatch between the ocfs and os 
>> currently used .  So we're confirming it with ocfs team. In the 
>> meantime, the customer is finding the update 1 OS so zhihui can do a 
>> testing on it.
>> xiaohui he wrote:
>>
>>> zhihui called me that such error disappear if putting them on raw 
>>> devices but still hang at the step when creating redo log.
>>> xiaohui he wrote:
>>>
>>>> How about set two raw devices for qurom and ocr  instead of
puting
>>>> them on ocfs partition?
>>>> zhihui yan wrote:
>>>>
>>>>> Hi Takeshi,
>>>>>
>>>>> Sorry for the late reply since it is difficult for me to
check
>>>>> mail in Intel.
>>>>>
>>>>> The error number is PRKR-1064 and PRKR-1005 during I
executed DBCA to
>>>>> create cluster database. I think all of them are resulted
from the
>>>>> problem that we
>>>>> can't create OCR successfully using "srvconfig
-init". In this
>>>>> case, gsd couldn't
>>>>> work normally. These two PRKR error are caused by this
problem.
>>>>>
>>>>> In this stage, I think there is one possible reason. That
is this
>>>>> version of OCFS
>>>>> for Itanium has one bug so that it couldn't support OCR
>>>>> successfully. Also maybe
>>>>> this problem could be related to bug2400133. I had reset up
one
>>>>> fresh environment
>>>>> for our continuous analyzation last week. Hope we could get
one
>>>>> reasonable and
>>>>> feasible fix this week.
>>>>>
>>>>> Thanks and best regards,
>>>>>
>>>>> Zhihui.
>>>>>
>>>>> Takeshi Watanabe wrote:
>>>>>
>>>>>> Zhihui,
>>>>>>
>>>>>>  
>>>>>>
>>>>>>> All those problems are resulted from PRKR or PRKC
error.
>>>>>>>   
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> What's the exact error number?
>>>>>>
>>>>>> If it's same as Intel's case, is it PRKR-1064,
right?
>>>>>> Note:212631.1 mentions several reason for this. Did you
check?
>>>>>> But PRKC error doesn't exist in the log Xihua sent.
>>>>>>
>>>>>> And I'd like to know test server's
ip/username/passwd in CDC.
>>>>>>
>>>>>> Regards
>>>>>> Takeshi
>>>>>>  
>>>>>>
>>>>
>>>>
>>>
>>>
>>

> First, we will get below error message when we issue "srvconfig
-init"
> command
> to initialize OCR.
checkk the bug databaase. this is an srv bug, been around for a long
time, there are patche at least for ia32. not an ocfs problem.
> Second, it will hang at 37% when we issue DBCA to create database. We found
> it hangs when it begin to create redo log files. Also need to ask OCFS 
> team for
> help.
hard to tell - we have this used in production so I suggest you do some
more debugging, good exercise to check stack dumps and see if dbca is
hanging inside dbca or inside a system call. a simple kernel stack dump
or ps aux and see if procesess are (and remain) in D state.