rdiff-backup doc in infra docs

2013-10-21 Thread Kevin Fenzi
Greetings. 

I just added a new doc to our infra-docs repo: 

http://infrastructure.fedoraproject.org/infra/docs/rdiff-backup.txt

Review/feedback welcome. Are there any unclear parts? Anything not
mentioned that would be good to know about backups?

Also, I'm going to do some restores from rdiff-backup in the coming
week and confirm all is looking ok. If it is, I am going to look at
stopping our bacula updates after Beta is out. Then we need to look at
some backup setup to backup our rdiff-backups (on netapp disk) to tape
(just in case). Suggestions for something to do those backups welcome. 

kevin


signature.asc
Description: PGP signature
___
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Re: rdiff-backup doc in infra docs

2013-10-21 Thread Matthew Miller
On Mon, Oct 21, 2013 at 11:23:19AM -0600, Kevin Fenzi wrote:
 Review/feedback welcome. Are there any unclear parts? Anything not
 mentioned that would be good to know about backups?

What about retention? How long are the backups kept?

On my own systems, I run rdiff-backup with `--exclude-if-present .NOBACKUP`.
This makes it skip any directories containing that file, and is useful for
subdirs with a lot of churn. For example, /etc/selinux/targeted -- but
perhaps most useful for people to put in home directories where they may be
dropping large temporary files that are actually okay to not be backed up.

Also, you probably already know this, but rdiff-backup does not have any
special handling for moved files -- they are treated as deleted and created
anew. This can mean that simply renaming something a lot can use up a lot of
space (unless of course the underlying filesystem is deduplicated, of
course).


-- 
Matthew Miller  ☁☁☁  Fedora Cloud Architect  ☁☁☁  mat...@fedoraproject.org
___
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Re: rdiff-backup doc in infra docs

2013-10-21 Thread Anshu Prateek
Few things
- its not mentioned that it runs rdiff as root. (guessing from sudo -i for
adding the key).
- guess its a pull backup than a push from the nodes being backed up?
- location of the backups on the filer?

regards
Anshu Prateek


On Mon, Oct 21, 2013 at 10:53 PM, Kevin Fenzi ke...@scrye.com wrote:

 Greetings.

 I just added a new doc to our infra-docs repo:

 http://infrastructure.fedoraproject.org/infra/docs/rdiff-backup.txt

 Review/feedback welcome. Are there any unclear parts? Anything not
 mentioned that would be good to know about backups?

 Also, I'm going to do some restores from rdiff-backup in the coming
 week and confirm all is looking ok. If it is, I am going to look at
 stopping our bacula updates after Beta is out. Then we need to look at
 some backup setup to backup our rdiff-backups (on netapp disk) to tape
 (just in case). Suggestions for something to do those backups welcome.

 kevin

 ___
 infrastructure mailing list
 infrastructure@lists.fedoraproject.org
 https://admin.fedoraproject.org/mailman/listinfo/infrastructure

___
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Re: rdiff-backup doc in infra docs

2013-10-21 Thread Kevin Fenzi
On Mon, 21 Oct 2013 14:06:06 -0400
Matthew Miller mat...@fedoraproject.org wrote:

 On Mon, Oct 21, 2013 at 11:23:19AM -0600, Kevin Fenzi wrote:
  Review/feedback welcome. Are there any unclear parts? Anything not
  mentioned that would be good to know about backups?
 
 What about retention? How long are the backups kept?

Currently forever. Sometime down the road we might start pruning them.
We are currently using about 2TB of a 25TB volume and it's growing
pretty slowly. 

https://admin.fedoraproject.org/collectd/bin/graph.cgi?hostname=backup03;plugin=df;type=df;type_instance=fedora_backups;begin=-31622400

I'll add a note. 

 On my own systems, I run rdiff-backup with
 `--exclude-if-present .NOBACKUP`. This makes it skip any directories
 containing that file, and is useful for subdirs with a lot of churn.
 For example, /etc/selinux/targeted -- but perhaps most useful for
 people to put in home directories where they may be dropping large
 temporary files that are actually okay to not be backed up.

Good idea, we could add this if there's call for it. 

 Also, you probably already know this, but rdiff-backup does not have
 any special handling for moved files -- they are treated as deleted
 and created anew. This can mean that simply renaming something a lot
 can use up a lot of space (unless of course the underlying filesystem
 is deduplicated, of course).

Yep. 

The netapp volume we are using is in fact deduplicated, also we aren't
backing up /var/log/ or the like on most machines. (we are backing up
logs from log02 which gets a copy of them, but it's setup to use
/MM/DD/ directories. 

kevin



signature.asc
Description: PGP signature
___
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Re: rdiff-backup doc in infra docs

2013-10-21 Thread Kevin Fenzi
On Mon, 21 Oct 2013 23:37:32 +0530
Anshu Prateek anshp...@gmail.com wrote:

 Few things
 - its not mentioned that it runs rdiff as root. (guessing from sudo
 -i for adding the key).

Yeah. Can add a note, yes it's root. 

 - guess its a pull backup than a push from the nodes being backed up?

Yep. It pulls from each client and does several in parallel. 

 - location of the backups on the filer?

/fedora_backups. Will add that too. 

 regards
 Anshu Prateek

Thanks. 

kevin



signature.asc
Description: PGP signature
___
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure