Copy from Datadomain to tape.. slow speed?

[athome]

Hi!

Does anyone have experience with copying from Datadomain (DD2500 10GB) to tape(LTO6) or from similar deduplication boxes?

I run a job that seems to be running at around 100 MB/s. I see that LTO6 should be able to run at 160 MB/s.

A

[dellock6]

Anders, do you mean a BackupCopy Job from the DataDomain to the LTO6?
Or a simple file copy between the two? In this case, what is the speed if you simply copy the files from the DD to another disk system? Could it be the DD the bottleneck nore than the tape drive?

Luca.

[athome]

Hi!

Yes I mean a backupcopy job from the datadomain to the LTO6. and yes it could be a CIFS/Datadomain-bottleneck-issue.

I also now see that the backupcopy-job copies ALL full backups.. so the backup to tapejobs just increase as the weeks go.. does anyone know if there is a way to limit it to only copy the latest full?

A

[Gostev]

The Backup to Tape job will only copy backup files that were not already copied to tape.

[athome]

The Backup to Tape job will only copy backup files that were not already copied to tape.

Is that really the way it works? Hmm.. There should be a option to only copy the LATEST full to tape. How else should one manage to differentiate on monthly/weekly to tape. I see that on the backup copy-job to repository.. you have alot of options to manage monthly/weekly etc.

Hope there is a patch coming with more options for the copy-to-tape-job.

A

[veremin]

Is that really the way it works?

Yep, only the data that hasn’t been backed up previously will be copied to tape medias.

How else should one manage to differentiate on monthly/weekly to tape.

Could you elaborate on it a little bit? Are you after tape GFS rotation scheme?

Hope there is a patch coming with more options for the copy-to-tape-job.

Since this was 1.0 version of our tape support, the existing functionality will be certainly enhanced/improved in the next releases.

Thanks.

[athome]

Could you elaborate on it a little bit? Are you after tape GFS rotation scheme?

Yes we would like to copy to tape on sundays(weekly) and the first of the month we want a copy (monthly)

A

[veremin]

Currently, GFS rotation scheme isn’t present in tape jobs, so, you have to use certain workarounds to make it work – this one, or that one. However, we do consider improving tape suport functionality in the next releases.

Thanks.

[athome]

Thank for replies

I think we will change our setup:
-Backup to local disk on the physical backupservers(daily incremental and full on saturdays)(only one week retention)
-Backup copy-jobs daily to copy data to DataDomain(incremental on weekdays, full on sundays)
-Tape backup of the full on sundays.

I assume we use deduplication and compression for the backups running to local-disk (with traditional incremental?)
and for the backup-copy-job to Datadomain we disable both deduplication and compression.

A

[veremin]

Please be aware that there is no “full” concept in case of backup copy job, since it doesn’t copy the existing backup files as a whole, but creates required restore points in target location, using its special "synthetic" logic. So, there won't be any "full" passes on Sundays, but "incremental" ones, instead.

As to your plan, it looks solid, and the statement regarding disabling compression is also correct.

Thanks.

[athome]

Hi!

I tried this concept now. of course backup to local disk works like a charm (about same speed as directly to the Datadomain ~300MB/s) but I think there is something about the copy jobs I don't fully understand. It is given a window to start at 06.00 in the morning.. and I see that it runs with good speed.. but it seems to stop at times.. and status go back to idle (in beteen some of the VMs I guess?).

I noticed it says in the log: 06.11.2013 13:16:24 :: Waiting for new restore points

Is this as expected? it still has about 15 VM-backups to copy over to the other repository.

A

[foggy]

Do these ~15 VMs have restore points within the specified "Copy every" interval?

[athome]

Thanks for reply.

Well all of the 68 servers were taken backup of from 20.00 to about 05.00 this morning. I now noticed that there where 24 VMs (not 15)..

The config for the copy job is "copy every 1 Day starting at 06.00

[foggy]

Anders, what if you click one of those VMs in the list to the left (in the job statistics window)? What do you see there? The fact is that if there is a fresh restore point (created within the copy interval) for this VM available, it should be copied by the backup copy job. If this is not the case, it is worth contacting support.

[athome]

Hi!

I now think I'm getting the hang of this. But one thing.. I have a backupjob to a localdisk-repository that only runs once a week (saturdays). How should a scheduele a backup copy-job that only runs once per week, and copies this to a Datadomain.

I tried Run every 1each day 05.00. and with a starting window only on saturday 05:00-16:00, this worked fine on saturday, but this morning I got an email: with this note on all of the vms: "Incremental copy was not processed during the copy interval"

[foggy]

How about setting the copy every interval to 7 days, so that the job does not attempt to run on other days?

[athome]

yes.. that was my thought aswell.. I was just unsure if I should set it to 6 or 7 . and what about the first time that I start it (ie today) will the job run on saturday?

A

[veremin]

Yep, it will.

If you specify the weekly synchronization interval (copy every 7 days) and start the job today, the job will be in the “idle” state during most days of week (Monday ->Saturday, Sunday -> Monday) and it will process the corresponding restore point (Saturday), once it appears. Also, in this scenario there will be no notifications sent about unprocessed increments.

Thanks.