Cloning and/or staging? 1

[litauer]

Please help with the following configuration:

I have a "big" LTO-library and a "small" file device. The file device is able to keep the backups of one month approx. The library should keep backups up to 6 month. I want my nightly backups go to the file device. But I don't trust the disks, so I immediatly want clones on tape. So far no problem. But ...
... as the browse and retention policies for a save set and its clone must be identical, I want the retention period to be 5-6 month, but have to set it to about 3 weeks because otherwise my file device will fill up. A dilemma? Any solutions? I read about staging, but this doesn't seem to be an option for me, as the save sets are removed from the file device after staging them to tape.

So for short, here is what I want:
- take nightly backups to file device,
- immediatly clone the save sets to tape,
- keep the save sets as long as possible (until it fills up) on file device for faster recoveries, and
- keep the save sets on tape for about 6 month.

Any ideas how to accomplish this? Thanks in advance.

Christoph

 

[Chacalinc]

- keep the save sets as long as possible (until it fills up) on file device for faster recoveries

before the disk fills up you need to remove the save sets, so staging is your solution! you have 2 modes, by oldest save set or space used. Read the manual about staging.

Cheers.

 

[litauer]

Well, I think I read the manuals ...

As you stated, I have two possibilies:

1.) stage oldest save sets if the disk fills up to a high water mark: In that case my save sets are only on the file device until it fills up to HWM. In my case this would last about approx. 4 weeks. So my disks have to be very reliable, because if they crash, I will lose the backups of about 4 weeks.

2.) stage save sets regarding the max. storage period: Because I don't trust my disks, I would set max. storage period to 1 day. After 1 day all my safe sets will be staged to tape -- and will be immediatly removed from disk. So I cannot do fast recover operations (from disk) of these save sets.

In fact I want kind of a "caching disk" that holds as much data as possible, but saves its contents to tape from time to time ...

 

[Chacalinc]

hmmm.. ok.. so did you think about using RAID 1 in your disk drives? at least if one disk drive fails you have the another one.

what is the OS of the NW server?

 

[litauer]

Yes, I thougt about ...
My file device in fact is a (software) stripe over two (hardware) RAID 5 "disks". Should be kind of reliable, but in the past we have had lots of trouble with the ide disks in the raids. So the chance of two disks of 12 crashing simultaneously is not too low ...

The NW server runs under SuSE Linux.

A possible solution: If I set the retention policy to 3 weeks, with backups to file device and cloning to tape. After 4 weeks, some of the tapes are marked recyclable, other tapes are still unused at this point. If networker asks for a new tape, which one would be labeled? An unused one or a recyclable one? If networker would recycle the "oldest" recyclable tape every time, my demands would be completed.

 

[Chacalinc]

If networker asks for a new tape, which one would be labeled? An unused one or a recyclable one?

1. marked as for recycle (no matter which one)
2. new tapes.

hmm... I'm thinking.. if you backup to disk (and cloning to tape) with a retention policy of the 6 months and you run a cron job (or schedule in case of windows) in order to remove the SSID in the disk drive older than 4 weeks? you can use:
(in unix/linux at least)
mminfo -r ssid -q volume="the_disk", savetime>4weeks > list.txt
cat list.txt | xargs nsrmm -d

 

[litauer]

This was a perfect idea, thank you very much! It solves nearly all my problems. The exact solution is:

mminfo -r ssid,cloneid -q 'volume=the_disk,savetime<-4 weeks,copies>1' |
tail +2 | sed 's-  -/-' |
xargs -t -n1 nsrmm -d -y -S

 

[Chacalinc]

that's cool!! glad to help you and thanks for post your script, it looks very good.

just a doubt, are you deleting the ssid AND the cline id?

Cheers.

 

[litauer]

just a doubt, are you deleting the ssid AND the cline id?

No, I don't think/hope so. As far as I understand the structure of networkers index, one save set can have more than one cloneid. One for each copy. If I would delete a ssid (as in your posting), all copies of this save set would be deleted. But I just want to delete copies of save sets residing on the_disk. nsrmm -d -S ssid/cloneid does (hopefully?) what I want.

 

[Chacalinc]

oh yes, I understand... my doubt was only by the "mminfo -r ssid,cloneid" command where you are getting the ssid AND cloneid and you pass that info to the "nsrmm -d" command.

I just didn't test the command and I'm sure you did.

Cheers.