Lukasz Szybalski wrote:
> In vmware server 1.0x ! Providing scripts would be great!
>
> - How do I create a snopshat via somekind of script?
> - How do I put "database into a consistent (sleep) state "
> - This will export "  vmdk-files" That get backed up by bacula? Correct?
> - Is that all?
> - How do I restore the whole vmware in case of system crash?
> - DB vmware needs to be shut down correct? (I saw some stop/start
> scripts on google, Is there something specific that they need to do

I only just noticed this one, but I backup whole VMs using bacula in the 
following fashion:

- VMWare machines are all stored in an LVM volume which is part of a VG 
with a certain amount of free space.
- All virtual machines in a subdirectory called FullBackupVMs are 
subject to being backed up in full.  The only "configuration" the script 
needs is that you need to ensure that any VMs which are to be fully 
backed up reside in a subdirectory of /home/vmware/FullBackupVMs and 
that all virtual disks they use are stored as files on the same LV.
- Broadly speaking, the backup process consists of:

1. Pause every VM.
2. Snapshot the logical volume.
3. Mount the LV snapshot
4. Restart every VM.
5. Let bacula backup the LV snapshot.
6. A post-backup script umounts and lvremove's the snapshot.

This has the minor drawback that every VM needs to be paused as part of 
the backup process - but the snapshots generated by VMWare server are 
binary deltas of things which have changed since the user clicked 
"snapshot" so unless the backup process itself is instantaneous and 
accounts for any data which may be in flight to the disk at the point 
you either run the backup or create an LV snapshot of the VM, there's no 
way you can guarantee you'll be able to go back to a snapshot. 

Naturally this does introduce some risk - you may come in in the morning 
and find that all your systems are in a paused state and have been for 
some time.  Myself, I only have systems stored in FullBackupVMs where 
there's little point in backing up only a handful of files because if 
the virtual server goes wrong it'll need to be restored in its entirety 
and where I can live with the virtual server being paused for some time.

Restoring is a case of "restore snapshot, import VM into VMWare, unpause 
VM".

A (slightly redacted) version of the pre-backup script which does steps 
1-4 above follows:


LV_SNAPSHOT=vmware-snapshot
LV=vmware
VG=sysvg
FULLBACKUPLOCATION="/home/vmware/FullBackupVMs"
MOUNTPOINT=/mnt
TMPDIR="/tmp/prebacula/$$.${RANDOM}"

mkdir -p ${TMPDIR}

# Sanity check:  does a snapshot VM already exist?  If so, something's
# gone wrong in the past and we should just drop out now rather than 
exacerbate
# the problem.

if [ -e /dev/${VG}/${LV_SNAPSHOT} ]
then
  echo "Snapshot already exists; failing"
  exit 1
fi


# pause VMs we want backed up.
# essentially, any VM which is running and lives in $FULLBACKUPLOCATION.

# sleep for 20 seconds between each VM pause because telling VMWare to pause
# too many VMs at once seems to upset it.

# List all VM's in $FULLBACKUPLOCATION

/usr/bin/vmware-cmd -l | \
        grep ^${FULLBACKUPLOCATION} > ${TMPDIR}/vmlist

# Find out state of each VM in $FULLBACKUPLOCATION, line number the result.

cat ${TMPDIR}/vmlist | \
        xargs -d'\n' -I"{}" vmware-cmd "{}" getstate | \
        awk '{print $3}' | \
        nl -s" " -w3 > ${TMPDIR}/vmstates

# Line number the list of VM's

nl -s" " -w3 ${TMPDIR}/vmlist > ${TMPDIR}/vmlist-numbered

# Join them, resulting in a list of VMs and their state.

join ${TMPDIR}/vmstates ${TMPDIR}/vmlist-numbered > 
${TMPDIR}/joined-vm-states

# Grep for the VMs which are running, turn the list into something usable.

grep "^.* on .*" ${TMPDIR}/joined-vm-states | \
        sed "s/^.* on //g" > ${TMPDIR}/working-list-VMs

# Create the scripts which will pause and restart each VM, based on the
# list of VMs which we want to stay on.

sed "s/^/vmware-cmd \"/g
s/$/\" suspend hard ; sleep 20 ; sync/g" ${TMPDIR}/working-list-VMs > 
${TMPDIR}/
suspend-VMs

sed "s/^/vmware-cmd \"/g
s/$/\" start ; sleep 20/g" ${TMPDIR}/working-list-VMs > 
${TMPDIR}/restart-VMs

. ${TMPDIR}/suspend-VMs

# snapshot /home
lvcreate -s -L6G -n ${LV_SNAPSHOT} /dev/${VG}/${LV}

if [ $? ]
then
  RETVAL=0
  mount /dev/${VG}/${LV_SNAPSHOT} ${MOUNTPOINT} -o ro
  if (! [ $? ] )
  then
    RETVAL=1
    echo "Couldn't mount snapshot /dev/${VG}/${LV_SNAPSHOT}"
  fi
else
  RETVAL=1
  echo "Couldn't create snapshot of /dev/${VG}/${LV}"
fi

# Regardless of how well that went, we'd better restart the VM.

. ${TMPDIR}/restart-VMs

# This is rather safer than an rm -rf of our temp dir

if [[ "${TMPDIR}" =~ "^/tmp*" ]]
then
  find ${TMPDIR} -type f -maxdepth 1 | xargs rm
  rmdir ${TMPDIR}
else
  echo "TMPDIR isn't in /tmp - leaving well alone"
fi

exit ${RETVAL}




-- 
James Cort

IT Manager
U4EA Technologies Ltd.


-- 
U4EA Technologies
http://www.u4eatech.com


-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users