• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu

JFS Hang and long CHKDSK after heavy I/O

Started by mobybrick, 2008.03.20, 00:02:55

Previous topic - Next topic

mobybrick

Dear all,

On one or two different systems, running an application that cause heavy intensive disk I/O, I am experiencing a hang. Thereafter, a full JFS CHKDSK is performed on my E volumes (JFS) that takes nearly an hour.

I seem to recall an ancient problem that heavy JFS I/O could cause a hang - but this problem still seems to be with us.

Anyone else seen this? Any suggestions? Will Serenity JFS likely have the same issue?

I am using IBM JFS 14.105

Thanks,
Moby

RobertM

Hi Moby...

I had a similar crash - but oddly a only few second longer chkdsk with a "the file system is clean" message (even though the crash was while writing to that disk).

One thing I read... someplace... said the default JFS settings are very unstable and should be replaced (by adding a cmdline switch) with:

/L:5,20,4

(to the end of the current JFS line).

As of yet, no crashes... but with or without the switch, writing to the JFS volume (at any decent speed) still seems to eat a lot more CPU resources than HPFS386 (on U160 SCSI no less) - but I need the large file support...

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Robert Deed

I can confirm that I have seen this under IBM JFS and I do not have this problem with openJFS (bootable jfs from RC4).  However I also made a hardware change which required me purchasing new memory, the older memory was being used on a development machine for a few weeks and proved to be faulty.  so I am not sure if it was crashing due to IBM JFS or if it was crashing due to bad memory (and possibly buggy cache management).

Quote from: mobybrick on 2008.03.20, 00:02:55
Dear all,

On one or two different systems, running an application that cause heavy intensive disk I/O, I am experiencing a hang. Thereafter, a full JFS CHKDSK is performed on my E volumes (JFS) that takes nearly an hour.

I seem to recall an ancient problem that heavy JFS I/O could cause a hang - but this problem still seems to be with us.

Anyone else seen this? Any suggestions? Will Serenity JFS likely have the same issue?

I am using IBM JFS 14.105

Thanks,
Moby

RobertM

Quote from: Robert Deed on 2008.03.20, 06:58:58
I can confirm that I have seen this under IBM JFS and I do not have this problem with openJFS (bootable jfs from RC4).  However I also made a hardware change which required me purchasing new memory, the older memory was being used on a development machine for a few weeks and proved to be faulty.  so I am not sure if it was crashing due to IBM JFS or if it was crashing due to bad memory (and possibly buggy cache management).

Which brings me to a few questions.... I'm not using JFS to boot from (nor plan on), but I do believe I need to update JFS. I have a version of IBM JFS from either 2004 or 2005 (latest one release).

I was considering updating to either the latest IBM JFS or Open JFS - but am very hesitant because I've got a few hundred gigs of files I dont want to lose (nor spend the time archiving - as many are used and locked by the server daemons and would require them to be shut down the entire time the files are being archived).

So my fear is, if I upgrade to the latest IBM JFS or to OpenJFS - can I expect to have data and partitions when I am done?

And the final question I had is... which is the better upgrade path? Performance wise? Stability wise? -and are there prerequisites (like the 1.104a kernel?) - and does the OpenJFS version have all the SMP concerns worked out? (I seem to remember that the early IBM JFS did not play well with SMP - even though it was supposed to utilize and benefit from SMP).


Thanks Robert or anyone else who has suggestions...
Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Robert Deed

I know the official word is that it may cause data loss on partitions which weren't formatted with the new system.  That being said I have 2 500GB drives which housed my MP3 collection in my old server.  The motherboard went south on me and I moved them to my workstation without thinking about it.  I have not experienced any data loss.  Perhaps someone from Serenity might chime in on this one, or maybe you should start a new topic so it can generate more feedback.

Quote from: RobertM on 2008.03.20, 07:07:51
Quote from: Robert Deed on 2008.03.20, 06:58:58
I can confirm that I have seen this under IBM JFS and I do not have this problem with openJFS (bootable jfs from RC4).  However I also made a hardware change which required me purchasing new memory, the older memory was being used on a development machine for a few weeks and proved to be faulty.  so I am not sure if it was crashing due to IBM JFS or if it was crashing due to bad memory (and possibly buggy cache management).

Which brings me to a few questions.... I'm not using JFS to boot from (nor plan on), but I do believe I need to update JFS. I have a version of IBM JFS from either 2004 or 2005 (latest one release).

I was considering updating to either the latest IBM JFS or Open JFS - but am very hesitant because I've got a few hundred gigs of files I dont want to lose (nor spend the time archiving - as many are used and locked by the server daemons and would require them to be shut down the entire time the files are being archived).

So my fear is, if I upgrade to the latest IBM JFS or to OpenJFS - can I expect to have data and partitions when I am done?

And the final question I had is... which is the better upgrade path? Performance wise? Stability wise? -and are there prerequisites (like the 1.104a kernel?) - and does the OpenJFS version have all the SMP concerns worked out? (I seem to remember that the early IBM JFS did not play well with SMP - even though it was supposed to utilize and benefit from SMP).


Thanks Robert or anyone else who has suggestions...
Robert

RobertM

Quote from: Robert Deed on 2008.03.20, 07:15:12
I know the official word is that it may cause data loss on partitions which weren't formatted with the new system.  That being said I have 2 500GB drives which housed my MP3 collection in my old server.  The motherboard went south on me and I moved them to my workstation without thinking about it.  I have not experienced any data loss.  Perhaps someone from Serenity might chime in on this one, or maybe you should start a new topic so it can generate more feedback.


Thanks! And done!

http://www.os2world.com/component/option,com_smf/Itemid,63/topic,852.0/

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|