VMware Cloud Community
Grove12
Contributor
Contributor
Jump to solution

RDM and VCB Performance

I have RDMs finally backing up via VCB without any issues... other than the performance is horrible. The file level back ups seem fine. But when I do a fullvm it seems to back up the entire RDM, regardless if there's data on it or not. I presented a 500GB lun as a test, put a few text files on it, launched a backup and it takes hours to complete. We're talking about a few kilobytes here. This happens whether I use the BEIM or the native vcbmounter... same results. It's not the SAN... performance is stellar other than this activity. Is this an expected behavior? Is VCB not smart enought to know if there's data on the RDM or not? I saw a few other posts about this, but never saw any resolutions. Just curious if you guys have any insights. There's gotta be a better way.

0 Kudos
1 Solution

Accepted Solutions
dconvery
Champion
Champion
Jump to solution

OK. By default, VCB will back up in a sparse file mode as long as you do not pass the -M or -F export flags are not passed on :

Export Flags:

-M: If set, the disk is exported into a single (monolithic) file.

When turned off (default), the disk is split into multiple 2GB files.

-F: If set, the disk is exported as "flat" disk, with no optimizations.

When turned off (default), the exported disk files will be more compact as

unused space in the disk image is not included in the exported file.

The BEST you will get with SAN mode FullVM type backups is about 1GB per minute. There is a further slow down with RDMs, because they need to be converted to VMDK during the FullVM backup. You would know this if you practice a full VM recovery (You ARE doing this right???). When you recover an RDM that is backed up using VCB, it becomes a VMDK.

RDMs are direct-attached LUNs in the OS native file system format (NTFS, ext3, etc) You may be better off using an agent or performing file level backups. Think about why you are using an RDM. If you recover to a VMDK, what are the implications. If there are non, convert to a VMDK and be done with it. If there ARE implications use file level or agent based backups. If these RDMs are for a file server, either method is fine. If they are mailboxes or databases, you are way better off doing agent based backups unless you REALLY know how to recover a potentially inconsistent database or mail store. The sync driver and VSS providers are not a guarantee that you will be able to recover.

Dave Convery

VMware vExpert 2009

http://www.dailyhypervisor.com

Careful. We don't want to learn from this.

Bill Watterson, "Calvin and Hobbes"

Dave Convery, VCDX-DCV #20 ** http://www.tech-tap.com ** http://twitter.com/dconvery ** "Careful. We don't want to learn from this." -Bill Watterson, "Calvin and Hobbes"

View solution in original post

0 Kudos
7 Replies
AndreTheGiant
Immortal
Immortal
Jump to solution

Which VCB transport method are you using? SAN?

Other VM with the same size use the same time?

Andrea

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
0 Kudos
Grove12
Contributor
Contributor
Jump to solution

I'm using San as the transport mode. I really don't want to backup over ethernet since we have 4GB fiber channel. Other VMs seem to back up much faster, but they are vmfs volumes. It's only the RDM that's an issue. It's a 500GB lun with around 20kb of data on it. I'd understand if it was full, but it's not even close. It seems that VCB can't determine that. No clue.

0 Kudos
AndreTheGiant
Immortal
Immortal
Jump to solution

I'm not sure that VCB can work in "compact" mode (download not the full vmdk, but only the data-trunk) also with RDM. I think that in RDM (obviosuly in virtual mode) will download the entire partition.

Andrea

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
0 Kudos
kjb007
Immortal
Immortal
Jump to solution

When you are backing up a fullvm, you are taking an image backup. So, no incremental or sparse backups in this method, you are taking a full copy of the vm, and then backing up that entier space that is occupied on the disk.

-KjB

VMware vExpert

vExpert/VCP/VCAP vmwise.com / @vmwise -KjB
0 Kudos
dconvery
Champion
Champion
Jump to solution

OK. By default, VCB will back up in a sparse file mode as long as you do not pass the -M or -F export flags are not passed on :

Export Flags:

-M: If set, the disk is exported into a single (monolithic) file.

When turned off (default), the disk is split into multiple 2GB files.

-F: If set, the disk is exported as "flat" disk, with no optimizations.

When turned off (default), the exported disk files will be more compact as

unused space in the disk image is not included in the exported file.

The BEST you will get with SAN mode FullVM type backups is about 1GB per minute. There is a further slow down with RDMs, because they need to be converted to VMDK during the FullVM backup. You would know this if you practice a full VM recovery (You ARE doing this right???). When you recover an RDM that is backed up using VCB, it becomes a VMDK.

RDMs are direct-attached LUNs in the OS native file system format (NTFS, ext3, etc) You may be better off using an agent or performing file level backups. Think about why you are using an RDM. If you recover to a VMDK, what are the implications. If there are non, convert to a VMDK and be done with it. If there ARE implications use file level or agent based backups. If these RDMs are for a file server, either method is fine. If they are mailboxes or databases, you are way better off doing agent based backups unless you REALLY know how to recover a potentially inconsistent database or mail store. The sync driver and VSS providers are not a guarantee that you will be able to recover.

Dave Convery

VMware vExpert 2009

http://www.dailyhypervisor.com

Careful. We don't want to learn from this.

Bill Watterson, "Calvin and Hobbes"

Dave Convery, VCDX-DCV #20 ** http://www.tech-tap.com ** http://twitter.com/dconvery ** "Careful. We don't want to learn from this." -Bill Watterson, "Calvin and Hobbes"
0 Kudos
Grove12
Contributor
Contributor
Jump to solution

The only reason we're using an rdm versus converting it to a vmdk to begin with is because of downtime. Re-attaching a lun via rdm takes about 1/100th the time as compared to converting 1.5 TB to vmdk. We dont get much downtime.

Thanks for all the responses. Extremely helpful. I guess it is what it is. I'll have to present my findings and see if the performance is acceptable. If not, perhaps we'll look at using some emc utilities like snapview to backup our rdms... or maybe file-level backups only. I really don't want to use an agent and send this stuff across ethernet... that's a lot of traffic. I had read before that the rdms become vmdks during restore. Not sure yet if that's an issue. Thanks again for the input.

0 Kudos
dconvery
Champion
Champion
Jump to solution

Oh yeah... Forgot about SAN based snapshots....

Dave Convery

VMware vExpert 2009

Careful. We don't want to learn from this.

Bill Watterson, "Calvin and Hobbes"

Dave Convery, VCDX-DCV #20 ** http://www.tech-tap.com ** http://twitter.com/dconvery ** "Careful. We don't want to learn from this." -Bill Watterson, "Calvin and Hobbes"
0 Kudos