Bug 75621

Summary:	Permission denied trying to touch a file; happens randomly; both ext3 and GPFS
Product:	File System	Reporter:	plvbr (phillip.l.viana)
Component:	ext3	Assignee:	fs_ext3 (fs_ext3)
Status:	RESOLVED UNREPRODUCIBLE
Severity:	normal	CC:	alan, phillip.l.viana
Priority:	P1
Hardware:	x86-64
OS:	Linux
Kernel Version:	3.0.13-0.27	Subsystem:
Regression:	No	Bisected commit-id:

Description plvbr 2014-05-06 18:45:11 UTC

Hi,

I am running a benchmark against DB2 in my cluster with several ext3 (SSD) and GPFS file systems. After running the benchmark for around two days I noticed that one of the nodes rebooted and that was due to an error touching a file. The file was being touched by one of the HA monitors of Tivoli System Automation.

I immediately tried to reproduce the error by touching the file again manually. I touched the file a few times successfully, then out of the blue I got a "Permission denied" error. Right after that I kept touching the file and got no error message.

CPU utilization and I/O utilization were NOT at their peaks. 

This error keeps happening randomly and therefore it is difficult to reproduce. It has happened both on a GPFS file system and an ext3 file system. Also I don't have much information on the internals of this problem.

Do you have suggestions on how to instrument/debug/monitor the error so that I can move on with the investigation and bring more info to this bug report?

Thanks

Phil

Comment 1 Alan 2014-05-19 11:08:06 UTC

Can you reproduce this on an upstream Linux system without GPFS, or does it need GPFS added ?

Comment 2 plvbr 2014-05-19 13:18:11 UTC

I couldn't reproduce it ever again. I reproduced it two or three times then it never happened. But as I said it happens on an ext3 system as well. Not sure if it would still happen on ext3 if I didn't have GPFS compiled.

Comment 3 Alan 2014-05-20 10:49:47 UTC

Thanks - given it was with 3rd party stuff loaded and not reproducible I'm not sure there is anything more we can do