can't figure out what I did

Brian J. Smith-Sweeney bsweeney at physics.ucsb.edu
Wed Dec 11 17:43:00 CST 2002


hello again all,

Well I'm embarassed.  Today I was messing about with the afacli to get
the feel for what it can do.  I had intended to experiment with changing
the chunk size on a RAID to see if it could be done without losing
data.  However, as I was looking at the docs I hit return when I meant
to hit backspace, and, well, I execute the command

    container reconfigure 0

Where "0" was the container number of the stripe part of my RAID 10.  

When I typed the above command the drives started working like mad and
the task at the bottom said it was reconfiguring.  I told the task to
stop quickly using "task stop".  Now the old container 0 seems to have
moved to container 60, and there's a new container 0 of type "Reconf". 
The old container is still of type "Stripe".  The full output of afacli
"container list" is now as follows:

AFA0> container list
Executing: container list
Num          Total  Oth Chunk          Scsi   Partition
Label Type   Size   Ctr Size   Usage   B:ID:L Offset:Size
----- ------ ------ --- ------ ------- ------ -------------
 0    Reconf 67.7GB            Open    
 /dev/sda             TEST-RAID10
60    Stripe 67.7GB       32KB None    
58    Mirror 33.8GB            None    0:00:0 64.0KB:33.8GB 
                                       0:03:0 64.0KB:33.8GB 
57    Mirror 33.8GB            None    0:01:0 64.0KB:33.8GB 
                                       0:04:0 64.0KB:33.8GB 

So my question is, what the heck did I do?  I found information on the
reconfigure command in the afacli manual, and it describes all the
things you can do by specifying additional flags like /chunk_size or
/volume.  But I didn't specify any of those.

The system still seems perfectly intact.  I tried to delete the new
container figuring that if it was live afacli would scream at me, which
it did claiming there was "important data" on there.  

Any ideas what happened?

Oh, and on a side note: I noticed some postings about recommended chunk
sizes earlier.  This systems going to be a mailserver (smtp/pop/imap)
using sendmail & UW-imap.  We have a mix of users; some have big
mailboxes, some divide things up into lots of little boxes.  I'm still
thinking thought the larger the chunk size the better as they'll most
likely be large sequential accesses.  Any opinions?

Thanks,
Brian "what's this big red button do?" Smith-Sweeney
-- 
========================================
Brian Smith-Sweeey
Senior Systems Administrator
University of California, Santa Barbara
Physics Department
bsweeney at physics.ucsb.edu





More information about the Linux-PowerEdge mailing list