Replication to new ESXi 4.1 server failing

[TheVirtualDude]

For the past week I have been testing offsite replication to our DR site using VBR 4.1.2 and HyperIP and I can't stop raving about the combination. However, the DR site server was running on ESX 4.0 and just yesterday I decided to upgrade the host server to ESXi 4.1. Now my replications are failing. I get the following error messages whenever the replication tries then fails:

Connecting Failed to login to "esx4.company.com" by SSH, port 22, user "root", elevateToRoot False, autoSudo False User authentication failed.

Also when I try to reselect the replica destination in the replication job I get the following error popup ""esx4.company.com" host uses different data transfer mode than the original host and cannot be selected.

I can successfully putty to the ESXi 4.1 host from the VBR server with no problems. I have tried re-registering the VM as well with no success. I am at a loss on where this could have broke.

Hoping someone has some ideas that I could try.

[Vitaliy S.]

Hello Faisal, you need to re-create your replication job, as switching between replication transport modes for "fat" ESX and ESXi hosts is not supported.

[TheVirtualDude]

Seems like that was it. I just recreated the backup jobs and its working now. But it appears that the transfer speeds are drastically slower on ESXi 4.1 than they were on ESX 4.0. Im using HyperIP as well and prior to the upgrade on some servers I could easily get speeds of 10-20Mb/s on replication jobs, now I'll be lucky if I hit 1Mb/s. This is painful. Any thoughts?

[Gostev]

It is expected to be slower. Replication to "fat" ESX is faster than to ESXi, because "fat" ESX allows us to use service console agents (which provides network traffic compression, and other benefits). However, 1MB/s is still too slow - I would expect to see at least 5-10MB/s (which is what I see in my our lab on cheap host with local storage), and seen 55MB/s reported on some nice storage configs on this very forums. Please try to upload a file to the same datastore using Datastore Browser in vSphere Client... do you get about same speed? Then probably ESXi does not "like" the host/storage/controller combination.

[tsightler]

1MB might not be too slow since he's replicating offsite. Replication with ESXi transfers significantly more data than replication with ESX so his performance will have more to do with link speed than anything else. What's the speed of the link in question?

[TheVirtualDude]

Thats a bummer. I really didn't want to run off of ESX alone since it seems like VMWare is heading towards the ESXi model. But it was truly amazing the replication speeds I was getting with ESX fat. I'm not sure if there are any tweaks or other adjustments that I could make that could fool Veeam into thinking its replicating to a fat model of ESX.

The link speed is a 10mb connection.

[Vitaliy S.]

Faisal, have you tried uploading any file to the same ESXi host as Anton has mentioned above? Along with that, I wonder if there are additional clues in the log files for your replication job, explaining why it is so slow.

[TheVirtualDude]

I haven't yet. I am absorbing the thought that Anton mention earlier regarding how Replication to "fat" ESX is faster than to ESXi, because "fat" ESX allows us to use service console agents. So currently I am upgrading my ESX 4.0 host to ESX 4.1. I am going to test again.

[Gostev]

It is simply impossible to use service console agents with ESXi due to its architecture, so there is no way to "fool" Veeam. Right now, "fat" ESX is a hands down better choice for replication target. And by the time ESX is removed from VMware offering, we will provide different architecture to better support ESXi targets. Thanks!