[OpenAFS] Pauses in vos dump commands

E. Chris Garrison ecgarris@iupui.edu
Thu, 13 Oct 2005 17:12:36 -0500

Dear OpenAFS-info folks,

Hello, I am new here.

I am getting my feet wet with vos dump backups for our prototype AFS
system.  I'm seeing some strange pauses in the backups though.

Here's what I do:

First, I do a "vos backupsys" for the server.
Next, I run a shell script that loops through all of the volumes on
the server, running the following:

  /usr/local/sbin/vos dump -id $i -time $since | <archive command>

Where $i is the name of the voluem, and $since is a string based on
the last full backup.

(don't worry about the archive command, it takes standard input and
saves data to tape across the network).

Most of the time, the vos dumps go quickly by, but every 4 or 5
volumes, it'll hang, seeming to think about the request a long time,
not even sending the first byte of data til it unhangs.  Usually this is 
around 10 minutes.  It doesn't matter if there were zero changes or a 
ton of changes, and it's not always the same volumes that hang.  
Once it unhangs, it zips through the dump and probably the next few 
as well, til it hits another one that makes it hang for no reason I 
can figure.  The load is 0.00 while it hangs, there I/O is doing
nothing... it's frustrating.

Has anyone else seen this behavior?  Should I have crossed my fingers
at the full moon while turning widdershins three times when I wrote
the script?  What am I doing wrong, or what might be causing this
hanging?  It could seriously slow down our backups when we get to
thousands of volumes to go through in a night.

Thanks for any help!

Chris Garrison			| Principal Mass Storage Specialist
ecgarris@iupui.edu		| Indiana University UITS - RAC - DSSG
W: 317-278-1207 M: 317-250-8649 | Jabber IM: ecgarris@itso.iu.edu