Unsolved
This post is more than 5 years old
33 Posts
0
2268
Confused about cloning from tape to tape and data amounts and tape numbers
We are using Networker 8.1, one server, OS is Windows 2012 R2. Networker does backups to Data Domain. At the end of week there is automatic clone job which clones weekend backups from DD to tapes. Tapes used for clone were 8 tapes and half of 9th. Beginning of new month I clone manually last week clone to new tapes as end of month backup. I am doing save set cloning from tape to tape through networker GUI. What confused me was that amount of tapes needed for month clone are 11 tapes even when there is 8.5 tapes amount of data in week clone. Tapes are brand new, first time in use, LTO4 and should be able to contain 800GB data and 1600GB compressed data. But somehow some month clone tapes contain less than 800GB data, one for example only 655GB. When I calculated the need for tapes I assumed that if I clone from tape to tape then I need same amount tapes, not more. I can only speculate that Networker clones some save set and starts cloning other and if it doens't fit to the tape then it marks tape full and starts writing it to next one?
Now, at the end of writing this and thinking about it made me realize that better way would be volume cloning. This would certainly use same amount of tape, wouldn't it? But anyway, it would be still good if someone would shed some light about tape to tape save set cloning peculiarities.
ble1
2 Intern
2 Intern
•
14.3K Posts
1
February 12th, 2015 14:00
Only way to have it same way would be volume duplicating and even then you might get some smaller differences (small enough to be ignorable). Because ssid may contain different parts on tape, depending on how compressible the part is, you may get different results. However, overall, if certain n number of ssids required m number of tapes, so would cloning. Difference may only arise if original set started to use beginning of tape and cloning end of tape which at the end of the day may give you picture you see. You can see this via mminfo by checking file and record numbers (file numbers would be relevant in this case). So, to get the same picture with cloning, you would need to emulate same picture as you have with source tapes which is pretty much impossible as we are not talking about repeatable patterns.
bingo.1
2.4K Posts
1
February 12th, 2015 14:00
In fact, NW does not really use "volume cloning" - it will clone "all save sets which are completely stored or at least have been started on the source media." Save sets will always be copied completely. So if one continues on another media, the whole
save set will be cloned.
Additionally you always clone to a pool, not to a certain media. The decision whether all save sets of a source tape will be stored on the destination media depends on the current scenario.
A third issue is that if you cloned a save set and the process has been aborted/failed, this save set will not be allowed to be assigned to the same media ... once retried, it must be stored on another tape.
Finally, extensive write errors could lead to the fact that the capacity of the destination volume is (much) smaller than the one of the source media.
In the end it is most unlikely that 2 tapes will have the exact same layout. But who really cares as long as NW knows where the save set can be found.
sasilik
33 Posts
0
February 13th, 2015 02:00
I used "volume cloning" because I can select volume and then select Clone command from menu.
And I think I can rule out errors because there would be something in logs if there were extensive write errors.
I estimated number of tapes for monthly clone to be same as number of tapes needed for last weekly clone when I clone the clone. But it wasn't and I try next time little different approach and think about it using thoughts from these couple answers.
ble1
2 Intern
2 Intern
•
14.3K Posts
0
February 13th, 2015 02:00
Bare in mind that volume cloning will clone ssids which start on that volume - middle parts and tails are not considered for individual clone cloning action. I think if you wish to estimate it, use following "algorithm"
- calculate total size for all ssids matching cloning action
- divide this average tape usage and round it to higher number
- compare this number to original tape number (backup tapes) - increase number if necessary (eg. more backup tapes)
- add one to two more tapes for extra buffer/eventual errors
bingo.1
2.4K Posts
0
February 13th, 2015 03:00
Keep in mind that you will not ask the media DB whether 2 copies of a volume exist (this is impossible with NW) but whether 2 copies/instances of the same save set exist.
If this is true, simply double their numbers and you are done.
BTW - with a Powershell script it is pretty easy to calculate the total of all save sets of a save set report ... as long as you use the same unit for the size (bytes, provided by mminfo's "totalsize" attribute).
ble1
2 Intern
2 Intern
•
14.3K Posts
0
February 25th, 2015 08:00
Could be up to the drive that you use for LTO4 drives on the box or even fw on the drives itself.
sasilik
33 Posts
0
February 25th, 2015 08:00
I went for the moment with solution where I run two identical cloning jobs in parallel. I got same amount of tapes for now. But I'll be damned if I understood this cloning thing. For example. I have one 2.1TB save set among others sitting on Data Domain. I clone multiple save sets to tape and this 2.1TB save set is among others. If I look how this save set is cloned to tapes then its like this:
tape1, head part, 468GB, parts of other save sets included on tape, total amount data on tape 823GB
tape2, middle part, 744GB, just middle part of 2.1TB save set, total amount data on tape 744GB
tape3, middle part, 669GB, just middle part of 2.1TB save set, total amount data on tape 669GB
tape4, tail part, 358GB, parts of other save sets included on tape, total amount data on tape 864GB and tape is appendable as its last tape of cloning job
Tapes are LTO 4, should be able to contain 800GB uncompressed data. Still don't understand why does it write only 669GB from the middle of save set and not more. As for other tapes, there is another 1.2TB save set where there is only 695GB head part on one tape(and tape is full with 695GB) and another part of save set is on another tape, 645GB + 53GB from another 770 GB save set and tape is full with 699GB. Can't see any logic in that. Files in save sets are mostly below 1GB, some are 1 or more GB but below 5GB. And no write errors in log files.
sasilik
33 Posts
0
February 25th, 2015 13:00
I guess I try out something if I have time. I take same 2.1TB save set and let Networker back it up directly from file server to tapes. Then recycle tapes and clone same save set from Data Domain to same tapes. And use same tape drive in both cases. My reasoning says that the amount of tapes should be almost same. But if backup job takes for example 2.5 tapes and clone job 3 and more then its mystery for me.