Slow (45 minutes) pause at the begining of replication?

[RichardR]

Hi All

Possibly a noob question so please bear with me.

I have Veeam Backup and replication 4.1 running between two WAN linked sites. The WAN is 100mbit with 38ms latency.

The replication is between 2 VMware 4.0 datacentres. I am using vStorage API SAN Only (working)
Network copy speed is good (25MBits per stream). CBT is not enabled as we are still using v4 hardware in the guests. (To be fixed....)

The problem is that starting the replication is very very slow when using large VM's.
We have a 560gb Exchange server and the replication takes about 45 to 50 minutes of "Searching" and "Creating Snapshot" before it actually starts copying any data.

Is this 'normal' or can we get it running any faster. We would like to replicate regularly, but this pause is actually one of the longest parts of the job! Once over the pause at the beginning we are getting "Backup speeds" (actually scanning speeds) of 100's of mb/sec (if there are not many changes to send over the WAN)

Cheers

Richard

[Vitaliy S.]

Hello Richard,

That's not an expected behavior, but it's hard to say what is the reason for that slowdown without seeing full logs. Have you already opened a support case? If not, please open a case with our support and provide the debug logs. For instructions refer to the support information provided when you click New Topic.

Thanks!

[nicko82]

Hi Richard, I think I may have experienced the same thing. We need to run a daily replication of a VM over a WAN Layer 2 link which is only 6M. I am using CBT so a 6M link is fine as the VM is fairly static.

My Veeam server was just another VM at our primary site with the source VM being replicated. The daily replications were taking on average 6 hours even though there were not many changes and CBT was enabled. I decided to to a simple test, I created a VM with 3GB vmdk I powered, I didn't install an OS I just left it powered on, and the BIOS was just searching for a boot device. NOW we definitely know there are no changes and the replications still took about 30 mins.

I decided to build a Veeam server however at the same site as the destination. I performed the same tests, replicating this test VM only took 2 mins instead of 30 mins. Now I run my Veeam server at the destination site for my production replication and using CBT the replications takes less than 1 hour. You see the Veeam server seems to do many tasks at the destination ESX server so I guess running the Veeam sever locally to the destination esx greatly improves the speed.

[RichardR]

Hi Nicko

That does look suspiciously like what is happening here, we have setup the Veeam backup servers at the source and copy to the remote over the WAN. This was so that we could use the direct SAN access at the source (fairly irrelevant I know as we only have 100mbit's to play with) but if it's there why not use it and without CBT on some of our machines it's the only way to go.

However I will test your theory and replicate one of the servers in the opposite direction (one that does have CBT) and see where that gets us.

If this is the case then I would prefer to run the backups with the backup server at the destination as this then gives us more options in the event of a catastrophic failure at the source as we will still have access to Veeam. Just have to make sure CBT is enabled, which is a reasonable job as we have 120 servers that are all in production .

If I can confirm this then I'll open a case to see if there is anyting Veeam can do to speed up the replication when there is reasonable latency.

Thanks for your help

Richard