Looking for a suggestion on best backup and copy method

[brianlees]

We are looking for suggestions on the best method for backup and copy in our situation. We are running Veeam 7.

Backup is currently to a Netgear ReadyNAS 516 on the local network. We have a second identical NAS (currently local) that will be placed in a remote office. We would like to replicate between them. I'm currently trying this via backup copy jobs.

We have four backup jobs:

• daily normal servers with a low retention requirement (14 retention points)
• daily long term servers with a much longer retention (30 retention points, 4 weekly, 12 monthly, 8 quarterly, 7 yearly)
• daily non-windows server with extremely low retention
• weekly server with extremely low retention

The thought is to backup to the local NAS, the replicate to the office after complete. So, we need to keep the deltas as small as possible. In addition, since the long term job has a pretty long retention, we need to keep the backup sizes as small as possible.

Based on an initial reading of the manual, it would sound like Reverse Incremental would be a good choice. But, I don't know how that will affect the replication job. Does it truly replicate only the deltas? Would it try to replicate the entire full?

[dellock6]

Local backups and remote backups via backup copy have nothing to share about backup methods. Local backup via reversed incremental is indeed the best way to reduce storage usage as much as possible, since only 1 full backup is kept. Be sure the NAS you are going to use is enough in terms of I/O to handle reversed incrementals, they are quite heavy on the backup storage.
About backup copy, they are always forward incrementals, and also the point in time you want to keep via GFS rotation are full backups too. You need to do some calculations about the space you are going to use on the remote backup. It's not a forever incremental.

Luca.

[brianlees]

Thanks for the info. The backups are already in motion, so I/O is not an issue. My bigger concern is reducing the amount being transferred over the WAN as much as possible. I read that the copy job is supposed to be only block level changes. However, the copy job says it has copied 69.3gb of data for the day...and we don't generate that much. I guess I'm going to have to look at the detail for each server to see where it is being consumed.

[veremin]

Do you see the comparable amount of changes with local backup jobs? Thanks.

[brianlees]

I do. I'm learning from reading that change block tracking includes the blank space on the drive? For example, we have terminal servers. Our users have roaming profiles, so they are deleted when the logout. Most people are logged out when the backups run, but it is still backing up 3.3gb of data in one example. The terminal servers are static and locked so it's only log files and profiles (copied over, deleted at logout) where the changed data is occurring. I have exclude swap file blocks (these term VMs have 22gb of ram assigned to them...so big swap files).

If Veeam is reading the blanks spaces, is there any way to say ignore that? From our view, if we delete it, it's deleted. I don't need to do a deleted file disk scan on a backup.

Of course, I'm new to Veeam, so I could be completely misinterpreting everything!

[veremin]

You, probably, have confused the ways file level and image level solutions work.

The situation you've described looks expected. VB&R is an image-based solution that cares only about blocks. If you write data to virtual disk and delete it afterwards, the corresponding blocks will be marked as changed ones. Then, these blocks will be backed up by VB&R.

Thanks.

[brianlees]

So, if I didn't want to copy the blank, but changed space on the drives, I would uncheck CBT. Will that make the backups less efficient? WIll it still only do the delta changes or will it copy entire files that have been changed? Basically, I want to make the backups as efficient as possible, both in speed and size, and make the WAN copy very small. Right now, our backups run very quickly...which is awesome. I could give up some time if I could still keep the size down.

Is it possible to also make a feature request to ignore change blocks that represent unused disk space?

[veremin]

The write/delete operations change the blocks. Veeam will use either VMware CBT mechanism or its own proprietary mechanism (if CBT fails for some reason) to identify changed blocks. Either way the blocks that have been identified as changed ones will be backed up by VB&R.

Hope this helps.
Thanks.

[brianlees]

If I disable CBT in the backups, will that make the backups less efficient? WIll it still only do the delta changes or will it copy entire files that have been changed? Also, how will that affect the copy job across the WAN? Will that still only do deltas or will it copy entire files again?

[veremin]

If I disable CBT in the backups, will that make the backups less efficient?

It will make the backup job longer, since the whole virtual machine image will be read.

WIll it still only do the delta changes or will it copy entire files that have been changed?

Only deltas.

Also, how will that affect the copy job across the WAN?

Whatever methods is used for identifying changed blocks in backup job, it won’t affect backup copy job.

Will that still only do deltas or will it copy entire files again?

Only deltas.

Additionally, it might be worth reviewing corresponding section at our Help Center.

Thanks.

[brianlees]

Perfect. That is exactly what I needed to know. Now, off to some testing!!!

[brianlees]

Is it possible to make a feature request to use CBT and have an advanced setting to ignore blocks for deleted files? If so, where to I send the feature request?

[dellock6]

The only way to know a block contains deleted files would be to read inside the Guest OS, but again Veeam is an image-level backup, not a file-level backup, and has no agent to be installed inside the VM...

Luca.

[yizhar]

For example, we have terminal servers. Our users have roaming profiles, so they are deleted when the logout.
Most people are logged out when the backups run, but it is still backing up 3.3gb of data in one example.
The terminal servers are static and locked so it's only log files and profiles (copied over, deleted at logout) where the changed data is occurring.

Hi.

You can reduce most of the roaming profiles changed blocks in several ways, such as:

* Use TS profiles instead of roaming profiles.
These profiles stays static in one location (file server). They are not synced to the TS local disk.

* Keep using roaming profiles, but avoid deleting and recreating them each time.
This will reduce changed blocks and probably also logon times.
Or - delete profiles only once a week/month instead of daily.

* Configure TS servers with second VMDK disk (drive D) and store user roaming profiles there.
You can then exclude those disks from backup.

* Configure separate VBR backup job for terminal servers, not with other servers such as FILE/DB/MAIL/AD .
This job will then have different (more relax) schedules, retention and replication settings then the other.

Yizhar

[veremin]

Is it possible to make a feature request to use CBT and have an advanced setting to ignore blocks for deleted files? If so, where to I send the feature request?

As mentioned, for now it’s not possible to implement this feature efficiently with image-level backup. Try to review the way profiles are currently set in order to reduce the amount of changed blocks, and, therefore, reduce the size of corresponding backup files.

Thanks.

[dellock6]

One future direction would be to use vSphere SE (space efficient) sparse disks and space reclaim operations from the Guest OS, but again it would be something done at the Guest OS level, Veeam would only have to recognize the reduced disk and act properly...

Luca.

[yizhar]

You can reduce most of the roaming profiles changed blocks in several ways, such as:

Yet one more option -
You can consider using "redirected folders" as alternative for roaming profiles, if desktop and my documents folder is what you actualy need to preserve.

[brianlees]

Thanks for the suggestion - we are using redirected folders in addition to the roaming profiles. Everything that can get redirected is being redirected. We just have so many logins and logouts to the servers that creates all those changed blocks. In addition, we have a few database servers that do dumps locally and then copy it to a different server. So, we have several GBs of data created and then deleted each day. I can probably change the jobs to copy directly to the server, but that would require discussions with the vendors since these are processes built into their software.

I love the concept of CBT, but it just isn't going to work for us given the amount of disk churn. Hopefully, some day, Veeam will be able to recognize a CB that is actually a deleted block and ignore it.

[brianlees]

Veeam 7

I need to setup our jobs to not use CBT (long story, other posts on it). In the backup job, I see two checkboxes in Advanced Settings/vSphere for CTB. Do I uncheck both to make sure that we are doing file level backups? Just want to make sure I get this setup right the first time.

Thanks,

Brian

[dellock6]

Brian,
at least for local dumps, one solution usually is to create a dedicated virtual disk, configure it as indipendent and persistent, and use it to save database dumps there. There is really no need to backup a backup disk, since you already have the same db in the running VM. This is one of the worst case for CBT, I've seen a customer for example with an Oracle Server, destroying and recreating a 300 Gb dump file every day. For CBT this is 300 Gb new blocks to be saved every day.

Luca.

[Vitaliy S.]

Brian,

Not sure I understand what you're trying to do. Veeam does VM image backups, no file level backups are possible. Moreover disabling CBT will not change the behavior of your backup jobs, as Veeam backup server will use our own change tracking mechanism to find VM changed blocks.

Thanks!

[brianlees]

I was told by Veeam support (can't find the email) that you can change the setting so we aren't backing up all the blank space that was churned over with file creation and deletions. That is our main problem. We have a lot of disk churn. HOwever, our backups are showing WAY more data being read than is actually being created. For example, one backup is showing 79gb being "Read". However, we are not generating anywhere near that level. MAYBE a gb and that is it. Part of that churn is from database dumps that get created, are copied to a different storage device, and then deleted. So, I don't want to back up all those changed blocks that are really deleted files.

Does that make sense? I understand exactly why this is happening, as they are changed blocks. But, since they are deleted files, we don't need those changed blocks backed up.

[brianlees]

I have already begun making other modifications to our VMs, such as moving MS Volume Shadow copies to different drives and excluding them. So, I can shave a little bit off, but the database dumps are a huge chunk of this.

[dellock6]

Yes, the size difference makes sense, as I said, consolidate dumps and VSS on one vmdk, and exclude that disk from backups.

Luca.

[brianlees]

So, what is the correct setting in the software to not backup all the individual blocks? Support said it was possible. The backup will take longer, since it needs to do file comparisons and save the deltas, but we should be fine there.

[Vitaliy S.]

I believe there was a misunderstanding between you and your support engineer, Veeam backup server does not compare files, it compares virtual disk blocks no matter whether you use VMware CBT or our own change tracking engine. Thank you!

[brianlees]

SO, is there no way to backup a VM without backing up all the "deleted disk space" CBs?

[Vitaliy S.]

Correct, as these blocks are treated as used/changed ones by VMware.

[brianlees]

Then what do the checkboxes about using CBT do if you can only back up the VM with CBT at the disk level and not the file level?

[veremin]

Check box determines what [url=hhttp://forums.veeam.com/viewtopic.php?f=2&t=18583#p91193]mechanism[/url] of changed block identification will be used – CBT or Veeam proprietary's one. Thanks.