07-20-2015 03:58 PM
We are running NetBackup 7.6.1.1. We have several remote sites. Each remote site has a local media server and MSDP. We have one central site with another MSDP. We use VMware policies with SLPs to backup data in the remote sites to respective media servers before duplicating the data to the central MSDP. Some of our servers have large amounts of data and our SLP windows and WAN links are not adequate to complete the initial/full duplication jobs.
While I have seen lots of discussions about seeding, most of them do not seem to apply or are quite complicated. I saw one discussion about using a Basic disk to seed the central MSDP but the details were lacking. What simple options can I use to seed our central MSDP?
Something similar to this??
1. Connect a USB drive and create a storage unit of type BasicDisk
2. Create a policy to backup data to the new storage unit
3. Execute policy
4. Detach USB drive and ship it to central site
5. Attach USB drive
Then what? Do I need to expire images? Do I need to import images?
While I am not new to NetBackup, I have never had to seed MSDP so any help is appreciated.
Solved! Go to Solution.
07-20-2015 09:37 PM - edited 02-12-2018 11:35 PM
To use BasicDisk, see Riaan's post in this discussion :
07-23-2015 02:07 AM
The idea behind BasicDisk using USB drive is just to have an easy way of transporting the backup to central site to duplicate to MSDP. This will serve as seeding from remote site and subsequent duplications from remote site will be optimised.
About explanation of Riaan's post:
Next step:
08-24-2015 08:00 PM
07-20-2015 05:03 PM
Is this a new environment? If not then I would assume that your central site would be adequately from the get go with data from its own backups. I have heard this discussion many times and the thought process hinges on the thinking at AIR or even MSDP to MSDP duplication is the same as deduplicating client to its MSDP. It is not. MSDP to MSDP through AIR or duplication is performed with a hash table that includes all the blocks of data in the pool. A deduplicating client deduplicates against is own image or group of images.
Only if your central MSDP is empty would you worry about seeding. If that is the case then we need a bit more information:
Do the remote sites have their own master servers thus making them their own domain?
07-20-2015 09:37 PM - edited 02-12-2018 11:35 PM
To use BasicDisk, see Riaan's post in this discussion :
07-21-2015 03:06 PM
Andrew, we recently replaced all of our backup servers and upgraded to 7.6.1.1 during the refresh. Our central MSDP has several terabytes of data in it, but it does not have images for 85 of our virtual machines in our central MSDP.
I have seen duplication jobs for larger virtual machines run for the same image for weeks. The duplication jobs always fail when our SLP window closes. One of our SLP windows spans 62 hours but the duplication jobs still do not complete in this time period.
We have a single master server with several media servers.
07-21-2015 03:08 PM
Marianne, Riaan's post is the one I referenced. I do not understand the step "Create an empty folder matching the remote storage unit and expire the backups". Can you explain this better?
07-22-2015 05:09 AM
Riaan's post is for a deduplication client to MSDP backup which does depend on the image being in place. Ray, your question (if I understand it correctly) is for MSDP to MSDP duplication which is not the same thing and should not be facilitated by this process. Are your remote media servers using deduplication pools?
07-22-2015 05:17 AM
Ray,
In the steps outlined by Riaan the expire is after the USB has been detached from the remote unit. The reasoning is that you do not want the images in the catalog or your import will fail. However I do not see where you stop the image cleanup from failing with a 1 from there on out unless you ship the USB back and reattach it to let the image cleanup happen normally.
07-22-2015 12:29 PM
After reading Riaan's post again, I agree that is does not apply to our situation. Sorry for the confusion.
Our remote media servers use deduplication pools. Backups from the remote VMs to the remote media servers/deduplication pools work just fine. Our issue is the duplication of images from remote deduplication pools to our central deduplication pool. Most of our VMs have duplicated images in the central deduplication pool but 85 or so do not. I have looked through Activity Monitor and can see where duplication jobs ran for the same image over and over. The duplication jobs never complete successfully. I assume they don't complete because the complete image is not duplicated before the SLP window closes.
I only have so much bandwidth and so many hours of the day to duplication images. I am looking for suggestions/solutions. Thanks!
07-23-2015 02:07 AM
The idea behind BasicDisk using USB drive is just to have an easy way of transporting the backup to central site to duplicate to MSDP. This will serve as seeding from remote site and subsequent duplications from remote site will be optimised.
About explanation of Riaan's post:
Next step:
08-24-2015 07:27 PM
We finally got a USB drive attached to the remote server and performed backups of several servers. We shipped the drive to our central site. The images were expired at the remote site. The images were imported at the central site. From what I can tell, the images are still on the USB drive and are accessible via the basic disk unit I created, but the images are not in our MSDP. How do I get the backup images into our central MSDP?
08-24-2015 08:00 PM
08-31-2015 09:19 AM
Thanks Marianne. I saw the duplicate option earlier, but the light bulb didn't go off.
I was able to import and duplicate several images. Some of the duplication jobs had pretty low deduplication rates while others had much higher rates than expected. However, when I activated images that had been suspended, the duplication jobs ran (over the WAN) for over 30 hours and failed to complete. I didn't expect the duplication jobs to finish immediate, but I didn't expect them to run for 30+ hours and not complete either.
Activity Monitor does not show the dedupe saving until a job completes successfully. Is there a way to tell if duplication jobs are taking advantage of the seeded images?
08-31-2015 10:22 AM
08-31-2015 10:47 AM
All of the backups are for virtual machines so client side dedupe is not an option without installing agents.
We have a mix of client names. Some of them match exactly while others do not. For example, most backups have the VM display name which is the NetBIOS name in uppercase. Most of the seeding backups have the fully qualified names in lowercase. That didn't seem to matter during the duplicate phase though. As noted early, NBU reportedly "sent" far less data to the MSDP than was initially backed up (9.26 TB backed up vs 4.40 TB CR sent).
08-31-2015 09:06 PM
09-01-2015 09:24 AM
Yes, CBT is enabled. Accelerator is enabled for the normal day to day backups. I did not use Accelerator for the seeding backups.
I suspended the largest of our images again to see if the smaller images would duplicate overnight. Two of the images did duplicate and show deduplication rates above 95%.
Another image started duplicating but appears to have had an issue. NBU shows the first 50GB chunk started processing at 8:47 PM and then nothing else was logged until the SLP window closed at 6 AM.
It looks like seeding worked. Thank you Marianne for your help!