Coda File System

Re: Amanda script to backup coda volumes?

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Fri, 2 Mar 2001 18:02:00 -0500
On Fri, Feb 16, 2001 at 11:54:41AM -0500, Brad Clements wrote:
> Was mentioned a while back, was wondering where I can find it..

I've placed Amanda + my current Coda changes under CVS. It is based on
Amanda 2.4.2 (+ debian patches). It might be possible to do some funky
CVS merging when you want to get rid of those debian changes which are
mostly for building the debian package and placing various files
according to debian policy (the filesystem hierarchy standard).

I've been running this against 3 of our servers on a daily basis. The
only amanda server change is to avoid dying on the new "CODA"
dumpprogram. Indexing of the contents of the dumps is not yet supported,
and what ends up on tape is a straight Coda volume dumps (no standard
formats like tar or dump). To restore such a Coda volume dump, pull the
dump from tape using amrestore, and then ship it off to a Coda server
using volutil restore.

Support for a CODA dumptype have been merged into the source for
selfcheck and sendbackup. For the size estimation there is a wrapper
script around calcsize which isn't installed automatically. After the
amanda-client installation (on the Coda servers) do,

    mv ${libexecdir}/calcsize ${libexecdir}/calcsize.real
    cp ${srcdir}/client-src/calcsize.coda-wrapper ${libexecdir}/calcsize

Where ${libexecdir} is something like /usr/lib/amanda.


Here is a short explanation of how to set things up,

Amanda.conf needs something like,

    define dumptype coda-sys {
	comment "Coda system volumes"
	program "CODA"
	options compress-fast
	priority low
    }
    define dumptype coda-usr {
	comment "Coda user volumes"
	program "CODA"
	options compress-fast
	priority high
    }
    define dumptype coda-usr-test {
	comment "Testrun for dumping Coda user volumes"
	program "CODA"
	options compress-fast, no-record
	priority high
    }

Then the disklist would contain entries such as,
    codaserver1   coda:root   coda-sys
    codaserver2   coda:root   coda-sys
    codaserver1   u:jaharkes  coda-usr
    codaserver2   u:jaharkes  coda-usr

And it is possible to still do regular dumps of normal disks,
    codaserver1 hda1 comp-root

When using a pre-5.3.13 server, the hexadecimal id of the underlying
replica must be used instead of the replicated volume names.

To get all this,

    cvs -d':pserver:anonymous_at_coda.cs.cmu.edu:/coda-src' login
    Password: <empty>

    cvs -d':pserver:anonymous_at_coda.cs.cmu.edu:/coda-src' co amanda

The Coda patches are on the trunk or HEAD branch. The vendor branch is
tagged AMANDA. On this branch are the pristine 2.4.2 sources, tagged as
V2_4_2, followed by V2_4_2_debian2 (the second debian release), etc.

To get just the Coda related diffs should be something like,

    cvs -d':pserver:anonymous_at_coda.cs.cmu.edu:/coda-src' rdiff -u \
	-rAMANDA -rHEAD amanda

Jan
Received on 2001-03-02 18:03:02