Offsite PBS with Tapes / Backup Strategy

[ITmarcapo]

Hi Everyone,
we are planning to expanse our backup strategy with an offsite PBS. We would transfer the data with syncjobs between the PBS.
The offsite PBS should additionaly backup these files on tape. At the moment we have no experience with tapelibraries. It would be awesome if you can check our strategy if we miss something important.

Library:
As library we would buy an Quantum SuperLoader 3 with LTO-9 drive / tapes and 2 magazine. Is the PBS smoothly working with Lib?

Backupjobs:
Our goal is to create multiple sets of tapes that we could store on an safe location. The data should be encrypted.
Our plan is to declaire 4 sets of tapes. So we would setup 4 different Backupjobs.

Our rotation / starting point will be the following:

1. SET A and B will be loaded
2. We start an manual Backupjob which copy all existing backups from all VMs to SET A
3. If finished we remove SET A from magazine and stored secretly
4. SET C will be loaded
5. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET B
6. 4 Weeks ago
7. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET C
8. SET B will be removed from magazine and stored secretly
9. SET D will be loaded
10. 4 Weeks ago
11. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET D
12. SET C will be removed from magazine and stored secretly
13. SET A will be loaded
14. 4 Weeks ago
15. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET A
16. SET D will be removed from magazine and stored secretly
17. SET B will be loaded
18. 4 Weeks ago
19. ...

With this rotation we minimize the problem that the copy process takes relatively long (over 24h) with one drive. The possible worst case time without an current backup on tape is only the duration that the initial copyprocess needs. Thats fine for us. And we only need to change the tapes every 4 weeks. Meanwhile we got an rentention up to 12 weeks for the tapes. The VM rentention is even higher because we copy not only the latest backup but all.

Does this strategy work like we thought it would?
Many thanks for your help!

 

[dcsapak]

As library we would buy an Quantum SuperLoader 3 with LTO-9 drive / tapes and 2 magazine. Is the PBS smoothly working with Lib?

i believe the model we have here to test mostly is a superloader3 but lto-8, but yes that should work

Our rotation / starting point will be the following:

1. SET A and B will be loaded
2. We start an manual Backupjob which copy all existing backups from all VMs to SET A
3. If finished we remove SET A from magazine and stored secretly
4. SET C will be loaded
5. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET B
6. 4 Weeks ago
7. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET C
8. SET B will be removed from magazine and stored secretly
9. SET D will be loaded
10. 4 Weeks ago
11. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET D
12. SET C will be removed from magazine and stored secretly
13. SET A will be loaded
14. 4 Weeks ago
15. We start an manual Backupjob which copy the next 4 weeks all existing backups from all VMs to tapes SET A
16. SET D will be removed from magazine and stored secretly
17. SET B will be loaded
18. 4 Weeks ago
19. ...

With this rotation we minimize the problem that the copy process takes relatively long (over 24h) with one drive. The possible worst case time without an current backup on tape is only the duration that the initial copyprocess needs. Thats fine for us. And we only need to change the tapes every 4 weeks. Meanwhile we got an rentention up to 12 weeks for the tapes. The VM rentention is even higher because we copy not only the latest backup but all.

mhmm... not really clear what you want to achieve with that.
please correct me if i'm wrong but AFAIU you want to do/have the following:

4 media sets with all backups of the past 4 weeks (relative to when the media set was exported/now)
2 are always in the lib
2 are always externally stored

am i right with this?

if yes, wouldn't it make more sense to just have one active media-set in the lib and two externally stored? (that way you save a whole media set)

i'd probably do a simple scheme for that with an allocation policy of 4 weeks (so that every 4 weeks a new meda-set gets started) with a retention of e.g. 8 weeks (so the finished media sets will be kept that long)

then after a media-set is fnished, export it and store it offsite

every 4 weeks (shortly before the allocaiton policy triggers), load the oldest media-set into the lib

or am i missing something here?

 

[ITmarcapo]

i believe the model we have here to test mostly is a superloader3 but lto-8, but yes that should work


mhmm... not really clear what you want to achieve with that.
please correct me if i'm wrong but AFAIU you want to do/have the following:

4 media sets with all backups of the past 4 weeks (relative to when the media set was exported/now)
2 are always in the lib
2 are always externally stored

am i right with this?

if yes, wouldn't it make more sense to just have one active media-set in the lib and two externally stored? (that way you save a whole media set)

i'd probably do a simple scheme for that with an allocation policy of 4 weeks (so that every 4 weeks a new meda-set gets started) with a retention of e.g. 8 weeks (so the finished media sets will be kept that long)

then after a media-set is fnished, export it and store it offsite

every 4 weeks (shortly before the allocaiton policy triggers), load the oldest media-set into the lib

or am i missing something here?

Thanks for your answer. Basically you are right.
Our problem is that we cannot 100% guarantee that "shortly before the allocaiton policy triggers". So thats the background with strictly defined periods.
Further at the moment we cant play with PBS and its Tapeoptions so we dont know how exactly we can avoid this problem with suitable configuration.
Because we dont need that many tapes this was the the safe option for us.

EDIT: Do you got any experience with an HPE StoreEver MSL3040 Library too?

 

[dcsapak]

Our problem is that we cannot 100% guarantee that "shortly before the allocaiton policy triggers".

what do you mean? why can't you do that? the allocation policy can be defined how you like, e.g. every first monday in a month at some specific time (or similar)

EDIT: Do you got any experience with an HPE StoreEver MSL3040 Library too?

no sorry

EDIT: also to play around with pbs and it's tape features you could use https://pbs.proxmox.com/wiki/index.php/Installing_a_Virtual_Tape_Library

 

[ITmarcapo]

EDIT: also to play around with pbs and it's tape features you could use https://pbs.proxmox.com/wiki/index.php/Installing_a_Virtual_Tape_Library

Thanks for the hint! Will definitive looking into it.

what do you mean? why can't you do that? the allocation policy can be defined how you like, e.g. every first monday in a month at some specific time (or similar)

Because the media-sets will be stored from different employees at various locations only accessible to them. So we must calculate illness or vacation.
Lets say mediaset from employee would the next to rotate. Unfortunately he is ill for 1 week. Out strategy would be then to take the current set 1 week longer in the lib and then go further in the correct order.