One of the more common questions I get around vSphere Data Protection (VDP) and VDP Advanced is whether backup to tape is supported. Before I answer that question and discuss a possible solution, I want to dig into why that question comes up fairly often.
As I am sure you know, tape has been around for decades for a variety of reasons. Some (not all) organizations who use tape typically have one or both of these requirements: Backup data must be kept offsite for disaster recovery and/or backup data must be archived for compliance reasons. Tape does a pretty good job of addressing both of these items, but tape can be cumbersome, costly, and unreliable. I managed a backup solution (I won’t say which one) back in the day. Every time someone would request a restore from older, offsite tape media, I would cringe. At that point, I knew a good chunk of my day would be spent retrieving the tapes, hoping the tapes actually contained the backup data I needed, and then performing the restore. Things worked as they should about 60% of the time. The other 40% consisted of troubleshooting, retrying the restore, recalling more tapes, etc. – often ending in failure. Now I suppose some of this could be blamed on the administrator (me), but I am guessing many of you have experienced similar pain.
I also did an informal survey in a session I hosted at VMworld in 2013. One of the questions asked was “How do you move backup data offsite for DR purposes?” Here were the possible responses along with the percentages showing how the audience answered:
“Send tapes offsite – we are happy with this.” 17%
“Send tapes offsite – there has to be a better way!” 26%
“Replicate backup data to another data center.” 35%
“Replicate backup data to a service provider/public cloud.” 13%
“We don’t send data offsite.” 0%
“Something else?” 9%
The audience represented a wide variety of organizations: small/medium sized businesses, education, government, enterprise and global organizations. As you can see from the numbers above, 17% of these customers are actually happy with a tape solution. 74% either wish to use something else or already use replication as the means to move backup data offsite. This is one of the main reasons VMware decided to invest in bringing backup data replication with VDP to market versus spending those R&D dollars on tape support. VDP and VDP Advanced can replicate to EMC Avamar (within an organization or to a service provider). VDP Advanced can also replicate to other VDP Advanced appliances. This backup data replication is extremely efficient – no need to pay extra for a WAN accelerator. Replicated backup data is encrypted for security. VDP Advanced also supports many replication topologies: one-to-one, many-to-one, and one-to many.
If backup data replication does not meet the requirements and tape is an absolute must-have requirement, that still begs the following questions: Does VDP and VDP Advanced include and support functionality for tape backup? The answer is no. Is there a fairly simple workaround to move VDP and VDP Advanced backup data to tape? YES! There are probably several ways, but here is how I did it:
I started by attaching an HP LTO Ultrium-6 tape drive to a Windows server and installed the HP StoreOpen Standalone version software. I formatted a tape with Linear Tape File System (LTFS), which gave me about 2.29TB of tape capacity. With these items installed and configured, I was able to simply map a Windows drive letter (G: in my lab) to the formatted tape. I can literally use Windows copy and paste commands to move files to and from this tape.
The backup data for a VDP appliance is stored in a set of .vmdk files (unless you use EMC Data Domain with VDP Advanced, which is discussion for another article another day). I created a short PowerCLI script to copy the important files that make up the VDP appliance from the vSphere datastore to the tape drive. Now I am no scripter/coder as I am sure some of you will quickly realize, but this set of lines got the job done:
I put the Start-Sleep line in there to give the VDP appliance plenty of time – 600 seconds, i.e., 10 minutes, in this case – to completely shut down. It can take several minutes for a VDP appliance to shut down gracefully. The files were eventually all copied to the tape, but this took a VERY long time.
How long? About two days! Granted, my lab hardware is aging so it is not the fastest environment. During those two days the VDP appliance was shut down so that I was copying a completely quiesced version of the appliance to the tape. That means I missed a couple of daily backups, which is not acceptable. I revised my plan to include two VDP Advanced appliances. VDPA01 became my primary appliance performing the daily backups. I used VDP Advanced replication to replicate the backup data from VDPA01 to VDPA02, my second appliance.
I replicate the backup data from VDPA01 to VDPA02 every Sunday morning. I have a scheduled task on my Windows server that starts the PowerCLI script a bit later on Sunday morning. VDPA02 is powered down for a long time while the copy is taking place, but that is ok since it only receives replicated data once per week. By taking this approach, I solved a couple of challenges:
No need to shut down my primary appliance – no missed daily backups.
Only backup data that must be sent to tape is replicated to the second appliance and copied to tape.
To verify this actually worked, I deleted the VDPA02 virtual appliance from its vSphere datastore, copied the files back from tape to the vSphere datastore, reregistered the appliance with vCenter, powered it on, and restored a VM from this appliance. While the process of copying the VDP appliance back from tape was time consuming, it all worked.
I realize this is a far-from-perfect solution. For organizations that need to archive backup data to tape once per week or once per month, it may be good enough, but doing this daily does not make sense. This process also assumes you do not have to retrieve data from tape often (hopefully, never!), but you still need to archive this data for compliance reasons. However, I would also like to point out some advantages of this method:
LTFS is supported by most if not all of the major vendors in the tape market such as HP, IBM, and Quantum. This article has more details on LTFS.
Specific backup and recovery software is not needed to recover the files that make up the VDP appliance from tape. There is no dependence on a catalog of backups/restore points to be available.
Using LTFS and .vmdk files will help lower risk when it comes to recovering very old data – these are standard formats that should withstand the test of time.
It is fairly simple and inexpensive to implement.
As for retention policies, you need to think about requirements, how often you perform restores, and how far back you typically need to go. For example, if…
80% of restores are from restore points that are less than 30 days old
18% are from restore points older than 30 days, but less than 90 days old
Remaining (2%) are from restore points older than 90 days
…then this retention policy below might be acceptable:
Again, I will be the first person to point out this is merely a workaround that will not meet everyone’s requirements. I simply wanted to at least share this work as it may be good enough for some.
Perhaps it is time to consider sunsetting those tapes in favor of a backup data replication solution! Challenge those that say “We are doing it this way because it is what we have always done…”
Now I just need to remember to change my tape every week. (Sigh)…I thought I was done changing backup tapes.
