This page describes the procedure for accessing and using our off-site backups. Only admins can do this -- if you want to get some file or directory back from the dead and are not an admin, please open a Bugzilla bug.

1. Backups of AFS Volumes

1.1. Navigating the available backups

Using backup-manager:

backup-manager list
backup-manager list YYYY.MM.DD

1.2. Retrieving a backup

(NOTE: $VOLNAME is not simply username, it is <db|mail|user>.USERNAME)

Using backup-manager:

backup-manager get YYYY.MM.DD $VOLNAME.dump.gz.aescrypt

1.3. Restoring the volume dump to a volume with a new name

Using backup-manager:

backup-manager restore YYYY.MM.DD $VOLNAME.dump.gz.aescrypt $VOLNAME.restored

Manually:

cat /vicepa/hcoop-backups/restored/YYYY.MM.DD-$VOLNAME.dump.gz.aescrypt | \
  ccrypt -cdk /etc/backup-encryption-key | \
  gunzip | \
  vos restore deleuze /vicepa $VOLNAME.restored

1.4. Mounting the newly restored volume onto the filesystem

fs mkm /afs/hcoop.net/.old/tmp-mount $VOLNAME.restored
vos release old

1.5. Restoring a particular file

# examine /afs/hcoop.net/.old/tmp-mount

1.6. Unmounting the restored volume

fs rm /afs/hcoop.net/.old/tmp-mount
vos release old

1.7. Renaming the restored volume so it takes the place of the damaged/corrupted/erased volume

Do this if you want to restore an entire volume. This deletes the old volume and replaces it with the backup.

vos remove $VOLNAME
vos rename $VOLNAME.restored $VOLNAME

1.8. Removing the restored volume

If you only wanted to restore a few files from the volume, you should remove the local copy of the backup volume when done.

vos remove -id $VOLNAME.restored

2. Database Backups

cd /vicepa/hcoop-backups/restored
mkdir YYYY.MM.DD-db
cd YYYY.MM.DD-db
cat ../YYYY.MM.DD-databases.tar.gz.aescrypt | \
  ccrypt -cdk /etc/backup-encryption-key | \
  gunzip | \
  tar -xvzf -

3. Proposal for New Backup System

by -- ClintonEbadi 2012-09-04 04:24:25

The current backup system has a serious deficiency in that it does a full volume backup every few days. This is untenable; we use ~4Mbit/s out of a 5Mbit/s allocation each month just for backups! More than ~150 members and we're toast. It also doesn't backup the local system data of any machines other than deleuze!

3.1. Requirements

• Encrypted backups
• Incremental backups
• Plays nicely with AFS

Thus, obnam. Things that might seem unobvious for anyone setting it up:

• afs backup volumes should be vos dumped (despite space waste locally) and backed up as a whole unit so that ACLs are preserved in the case of restoration

3.2. Basic setup

Every day,

• Pull system / database backups

• vos backupsys (remember to ensure all are mounted in /afs/hcoop.net/old/ for member benefit, and vos release the volume)

• dump all backup volumes
• sync with remote obnam repository

Bandwidth permitting, of course. Store as much as we can remotely; 30 daily backups and another month or two of weekly backups seems reasonably sane.

To secure the repository, we need a backup afs user (possibly exists already) that can read a passwordless gpg key (and ssh key for logging into rsync.net) securely stored in afs. The backup user will need a keytab, but it needs to be propagated to only the needed machines (perhaps time to clean up our keytab syncing regime in general?). Each active admin should generate a gpg key specifically for the backup repository (good for a limited time, perhaps one year) that will then be added so they can access backups on behalf of members. When admins depart, their keys can easily be revoked, unless they were malicious and stole the obnam key.

Initially, just live with backing up user volumes (far, far better than nothing, and we have no disaster recovery plan for using system backups anyway so their value initially is dubious, since they've effectively never been supported). Back up databases first, and then local systems (user data is more important -- once we rid ourselves of deleuze in theory machines should have no unique data of importance).

3.3. Incremental implementation

To make life simpler, individual machines would be backed up to afs volumes ({recovery|backup}.$machine.hcoop.net?). Probably as uncompressed tarballs, using an adaptation of the current system backup scripts to only copy files that cannot be restored automatically by setting the status file and running apt. Should each machine have a unique backup user (what happens if you have a $user/$host.hcoop.net key? Can it authenticate on all nodes via pam_krb5, or just on $host?)?

Databases could be backed up similarly, just by rsyncing over the /srv/databases directory. Same issues as previous backups with possibly bad file system state could arise however... if it's not impossible, we should do a proper database dump for each database separately. Perhaps we could "resurrect" $user.db, but instead of $user.dbbackup containing the last snapshot of the user's database dumps? Possibly not worth the effort / it might be better to keep them unexposed to the world at large (violation of database acls).

spamassassin should probably be writing its bayes database directly to afs anyway, so we can punt on that. There should not be any other local state.

In a world where every child gets a free pony for their tenth birthday, we could teach obnam about afs acls and just mount the $user.backup volumes (i.e. not double the local space requirements for volumes!) and backup from those with the intermediate dump. This would also allow users to control what data gets backed up via ACLs. However, reality bites.

CategorySystemAdministration