separate backup network

[renebos]

#5185811

I've created a seperate backupnetwork within my vsphere-cluster with 5 ESX hosts seperate from managment, lan , dmz and vmotion ,network called "backup".

This network from each host is attached to G8000 Switch (1Gbit) and then connected with 10 Gb Emulex to Server with windows 2008 R2 as OS.

Vcenter has 2 networkcards in windows, one with internal lan and one with backupnetwork.

I added VC in VEAAM I see it from server through backupnetwork. I then make backupjob and VC makes snapshot, however it is tranferring data through LANNETWORK and not BACKUP.

VEAAM SERVER is also connected to LAN and BACKUP network.

How to make the server backup over the seperated network instead of LANNETWORK?

I also run Tivoli Storage Manager on the server so i need my LAN CONNECTION for TSM-backups.

Does the above make any sense to you guys out there?

[renebos]

When I remove the LAN-connection the backup also failes.

Is there any rule within vmware or veeam that decides what NIC the backup gets tranferred over?

[dellock6]

Is vCenter listening on the backup-net IP? Have you registered vCenter inside Veeam with that IP? Try first to check those two.

[renebos]

Tnx for your reply, yes

vcenter has a second NIC adressed in the backupnetwork and is pingable from veeamserver and i can ping backupserver from vc also.

[renebos]

It is also added as backupip in veeam

[dellock6]

Ok.
Which kind of storage do you use, and how do you reach it from a Veeam perspective? Is it done via network mode (I suppose) ? Since Veeam is a physical machine, you can design the backup infrastructure to let Veeam acces directly the storage, and then using the connection with vCenter (no matter via which network) only to request snapshots activities to vCenter itself.

[rgirard-wwf]

Makes sense to me.... Veeam will communicate with your hosts directly. You'll need to add the backup network to your ESX hosts and assign vmkernel ports on the backup network. On Veeam, make sure dns names for your ESX hosts resolve to the ip addresses you assigned to your vmkernel ports that are on the backup network.

Hope this helps,
Rob

[renebos]

Ok.
Which kind of storage do you use, and how do you reach it from a Veeam perspective? Is it done via network mode (I suppose) ? Since Veeam is a physical machine, you can design the backup infrastructure to let Veeam acces directly the storage, and then using the connection with vCenter (no matter via which network) only to request snapshots activities to vCenter itself.

For ESX we use EMC CX4 as storaga FC connected on the hosts.
For Veeam as backuprepository we use SAS connected storage.

Veaam does not have a FC connected to ESX-SAN, maybe that is a wise idea to do?

[renebos]

Makes sense to me.... Veeam will communicate with your hosts directly. You'll need to add the backup network to your ESX hosts and assign vmkernel ports on the backup network. On Veeam, make sure dns names for your ESX hosts resolve to the ip addresses you assigned to your vmkernel ports that are on the backup network.

Hope this helps,
Rob

You hit the nail right on its head.

Thanks i did what you suggested DNSWise, so i configured the esxhosts in hostsfile on the veaamserver and backup went through backupnetwork.

Its backupping with 49 MB, with bottleneck is source 99%, proxy 45%, Network 0%, Target 0%, this is on a seperate Gbit network. Is that acceptable?

Or is this resolvable by putting FC HBA in server (if there is still room left for an extra card) and connect is directly to ESX-SAN.

I also have the option to trunk several NIC's, would that speed up transfer?

Any help still welcome.

[dellock6]

Hi Rene,
you have to tell us if that speed is acceptable
Anyway, coming from e EMC CX4 49 mbits are quite low, check in the logs, you are using network mode as I asked before?
If so, yes having FC HBA on the Veeam server and using SAN mode will almost surely increase performances.

[renebos]

yes, NETWORKMODE is beeing used at this time.

[dellock6]

Uhm, ok, so that speed is somewhat to be expected, network mode is the lowest method.
If you can go for FC HBAs on the Veeam proxies you will for sure see an increase in the backup speed, since I'm sure the EMC storage can do much more than 49 mbits, even without knowing which kind of disks and raid are there.

[renebos]

Now it is getting a bit tricky for me.

I took a server put in 2 HBA's connected them to FC-switch, windows 2008 installed.

On the FCSwitch

zoning proxy01 added.

On the EMC-San:

i have my ESX Hosts in Storagegroup_ESX with attached LUNS.

Should i put my proxy01 into the same Storagegroup_ESX, then it will be attached to the same luns the ESXdatastores are, but can the windowsbox do anything with those luns.

Could windows mess up my datastores?

What should i do?

[dellock6]

Veeam disable the automount features of Windows so you will be safe, but if you want extra security and if your storage allows it (I do not know anything about EMC), expose the luns to Veeam proxies in read only mode.

[renebos]

There are some readonly options, depends on the softwareversion of the EMC. Need to figure this out without corrupting data.

For people with EMC knowledge.

I'm trying to create a 'BACKUP-Storagegroup' that only has readonly access to the same luns that are in the 'ESX_Storagegroup' where my ESXHOSTS are in.

Do i need to install emc powerpath 5.1 on the proxy?

Please advice me how to let the proxy access the same luns that my ESX uses for the datastore without corrupting the data.

[renebos]

Either the SAN CX4 or Fabric DS-300B don't provide the readonly options so support told me.

There seem to be some option to get a storagegroup readonly, but only through cli.

[got this from other forum]

Can you add a LUN as read only? - Yes, if you are running Flare R31.5.5xx or greater. It can only be done from the cli and the command is: naviseccli -h ip storagegroup -addhlu -readonly

Since i don't want to experiment on a live situation i did the following.

from command console

Code: Select all

diskpart

automount disable [prevents automount of new volumes]
automount scrub [clears previous mounted volumes]

and in registry HKCU->Software->Microsoft->MMC-->{8EAD3A12-B2C1-11d0-83AA-00A0C92C9D5D} add REG_SZ Restrict_Run with value 1

now the Disk Management isn't available anymore through normal ways prevents users to think twice if they want to do anything with the disks.

I will post the speed i am getting now in a later stadium.

[renebos]

CAN'T BACKUP ANYTHING NOW.

What did i do wrong?

vmname Error 4:08:42 PM 4:11:59 PM 47.3 GB 0.0 KB 0.0 KB 0:03:17 Error: Client error: NFC storage connection is unavailable. Storage: [stg:datastore-21,nfchost:host-954,conn:192.168.99.253]. Storage display name: [EMC Datastore 2 (1TB)]. Failed to create NFC download stream. NFC path: [nfc://conn:192.168.99.253,nfchost:host-954,stg:datastore-21@vmname/vmname.vmx].

[renebos]

Enter names and ipadresses of backupnetwork in proxyhostfile.
Opened 902 port in firewall eventhough it was disabled.

Running AT 34 MB/s

Bottleneck proxy

95% = 99% = 31 % = 2 %

CPU load is 100% on this quadcore CPU, maybe i should add CPU's in this pysical machine.

this thread can be closed!