Bug 58440

From Bugzilla Helper:
User-Agent: Mozilla/4.78 [en] (Win98; U)

Description of problem:
I have been using the dump program to backup my root filesystem to a disk file.
This is a technique I did nightly on my Redhat 6.2 system without 
problem. Now on Redhat 7.1 the dump of root routinely hangs at various points.
I've seen the hang at 350Meg, 750Meg, and 1.1Gig. Yet I have also seen 
the dump run to completion.

In each case where the hang occurs there are no messages produced and there is
plenty of disk remaining. No further bytes are written to the 
output file. There is usually 1 parent dump process and 3 children shown in the
process list, none of which using resources. Doing a "C" causes 
the dump process to prompt the user if the dump should really be terminated.
Answering "yes" causes all processes to be cleaned up normally. I have 
allowed the dump to sit for as long as 1 1/2 days and it never completes.

This failure occurs when the output file is on an internal disk filesystem (SCSI), or
an external NFS filesystem.

A fix would be greatly appreciated as this bug causes problems with automation
and is a concern for our production environment.


Version-Release number of selected component (if applicable):


How reproducible:
Sometimes

Steps to Reproduce:
1. dump -0auf /mnt/root-full.dump -L "FULL`date +%Y-%m-%d`" /
2. wait for failure
3.occurs about 75% of the time.
	

Actual Results:  The following output is produced:
DUMP: Date of this level 0 dump: Wed Jan 16 09:50:03 2002
DUMP: Date of last level 0 dump: the epoch               
DUMP: Dumping /dev/sda1 (/) to /mnt/root-full.dump       
DUMP: Label: FULL2002-01-16                              
DUMP: mapping (Pass I) [regular files]                   
DUMP: mapping (Pass II) [directories]                    
DUMP: estimated 1109835 tape blocks.                     
DUMP: Volume 1 started at: Wed Jan 16 09:50:04 2002      
DUMP: dumping (Pass III) [directories]                   
DUMP: dumping (Pass IV) [regular files]

The output file continues to grow for a while and then stops. Little to no CPU is used by the dump processes. I've let it sit for up
to 1.5 days with no additional bytes posted to the output file. A <cntrl>C will issue prompt and ask if I'm sure I want to kill the dump.

Expected Results:  Instead of a hang, the following additional messages should be issued upon completion:
DUMP: 100.00% done at 3822 KB/s, finished in 0:00        
DUMP: Closing /mnt/root-full.dump                        
DUMP: Volume 1 completed at: Wed Jan 16 09:55:23 2002    
DUMP: Volume 1 took 0:05:19                              
DUMP: Volume 1 transfer rate: 3804 KB/s                  
DUMP: 1213695 tape blocks (1185.25MB) on 1 volume(s)     
DUMP: finished in 319 seconds, throughput 3804 KBytes/sec
DUMP: level 0 dump on Wed Jan 16 09:50:03 2002           
DUMP: Date of this level 0 dump: Wed Jan 16 09:50:03 2002
DUMP: Date this dump completed:  Wed Jan 16 09:55:23 2002
DUMP: Average transfer rate: 3804 KB/s                   
DUMP: DUMP IS DONE


Additional info:

Here, Redhat 7.1 is running on an IBM x330 Server with dual Pentium 3 - 1 Gigahertz
processors, 1 gig ram, dual 9 gig internal disk (Raid mirror) where the OS 
runs. Kernel is 2.4.9-12smp. Dump program shows version 0.4b21-3.