I was talking about the configurations of the bacula user that originated this thread. And I based my opinion on his pool configuration and list media output.
Well, in my very humble opinion there is no any evidence that RAT or Dimitri is making Full everyday.
I agree with you that some files are always backed up regardless of whether or not you are using a full backup job. And this is the case of virtual machine disk images.
You can always use a Delta plugin, Bacula GED(tm) or for VMware a Bacula vSphere plugin to get an block level Incremental backup of virtual machine images. You can save a lot of space in your archive. Delta and VSP is also working with tapes.
We have XenServer virtual machines. AFAIK even the enterprise version do not have this feature yet.
Well. I do not agree. You can use a Delta plugin or Bacula GED(tm) with XenServer too. Yes, BEE do not have a dedicated Xen plugin yet, but you can make a virtual machine backups without a dedicated plugin. Or you should write a Xen plugin for Bacula. This is an OpenSource project, so when someone wants a functionality which is not available then he/she should write it yourself.
I needed a pgsql plugin for Bacula, I developed it. I needed a AD/LDAP object level backup and restore with Bacula, I developed it. I needed block level deduplication in Bacula, I developed it. Now I need some functionality in BWeb, so I'm developing it.
.
I have seen solutions that have a weekly or monthly full backups for VM disk images and daily backup for the data partitions of the virtual machines that are susceptible of changes dieting the week/month.
Sounds to me like a tricky to manage.
Most of our virtual machines do not need frequently changes (remembering that data - log and conf files - is regularly backed up). We have upgrade plans and try to implement them immediatly after a regular full backup.
I was talked with a guy who manage a few thousand virtual machines environment at some company. This machines are used for web application serve. And they do not backup them at all. First an average lifetime of the virtual machine in this company is about 2H. So there is no time to backup. An old machine is disabled and deleted, then a new machine is created, powered on, installed, configured and enabled. And so on, every day, every hour. When they wants more computing power to handle more throughput, then they rise a number of new vm created in time. It was Off-Topic, just a curiosity.
Also, the most cases I had seen that needs a Virtual machine disk image restore are:
1) disaster recover: in this case, the last backup is what we need. Having this backup in more than one place is preferable than having later ones.
Right. Then you can use a job replication (SD->SD Copy Jobs) and techniques mentioned above for optimal performance.
yes :)
2) updates and/or upgrades in the virtual machine configuration do not work: in this case we need a backup immediately before the changes were made. In this case we can take care of always having a full backup before doing any software changes in the virtual machine.
I do not understand what you mean. Sorry.
I mean that if an update and/or upgrade of any of your virtual machine software (OS, app, etc.) do not work, you will need to restore from a backup immediatly before the changes were made. This could be achieved restoring the last full virtual machine backup + the last full/incr/diff data backup (logs, cfg, etc.).
As Dimitri mentioned before, for this kind of purpose you should use snapshots, not backups. Backup only when your snapshot mechanism is not working and you lost data. When I had to restore virtual machine in this case I'll change a virtual hypervisor next day.
