cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Announcements
1279
Views
0
Helpful
4
Replies
Highlighted
Beginner

2 x NSS322 - iSCSI replication

Hi

My plan is to create an iscsi datastore for two VMware ESX 4.1 U1 servers (HP ML110 G6). The iSCSI target should have a redundancy for a failover in the case of a h/w failure.

Is it possible to replicate iSCSI targets between two NSS322? ..maybe on block level?

Tnx for your suggestions..

4 REPLIES 4
Highlighted
Cisco Employee

To answer your question directly, yes this is possible. BUT!

I do not think this is really what you want to do. What I would recommend is to look into Multi-Path solutions from the NSS to the ESX server, which will give you the redundency you are looking for. Then to "back-up" your iSCSI target you can configure the NSS's for rSync, one to the other. The destination should not be accessible by the ESX server as that will be only for iSCSI image backup file.

If you are really familiar with Linux, there are other copy functions you can use, like 'dd' for example, but rSync would be the easiest.

Hope that helps.

Highlighted

Hello Alejandro

Thanks for your answer.

How would it be possible to switch to the second NSS in case of a h/w failure if it is not reachable by the ESX's? In your explanation, is it intended as a cold standby?

The NSS integrated rsync function is working with a scheduler. I am looking for a solution with a continuous replication. Veeam Backup & Replication version 5 has a new feature named "SmartCDP" for near-continuous replication. Maybe this would be a solution...

Highlighted

Let me answer like this,

Since to our client (ESX server in our case) this is an internal HDD we need to start thinking first an foremost how are we going to back up the server. This is very different from how are we going to back up the iSCSI target. So using an utlity like veeam is perfect, but we are backing up the data not the iSCSI image contained in our NSS. If you need to recover your VMs from any point in time then sounds like you are looking in the right spot.

What I meant about not allowing the ESX server see the other NSS is that we do not want the ESX server to try to mount and write to our image in two different places. It should only look for its iSCSI target on one NSS only. Now, by using rSync or 'dd' you can remount the same image if something goes downhill fairly quick. Granted you will have to remap the image but if you are careful I have been able to do this with no issues. Keep in mind that this is not supported and in my case it was for testing and all my servers (3 2008r2 x64 and 1 2003 i386) were for testing not production. The NSS is perfectly built and capeable of using MC/S and MPIO which is what I would be more concerned about; a network failure. My redundency is focused around this (network connectivity) first then for critical server we use vCenter Server to control our cluster of 6 ESX servers. vCenter handles our fail-over of VMs should something go wrong.

The bottom line is that we cannot connect to a single HDD that is in two different locations. When we look at it in this perspective it is simple to see how trying to write would be a difficult operation.

This iSCSI target(s) are they for datastore or are they raw maps for your VMs?

Highlighted

I use both, iSCSI targets with VMFS and different iSCSI targets which are raw device mapped (Symantec Backup Exec server).

A separate NSS as a "passive" failover would be nice, but i'm searching for a solution with a continuous replication, or "near"-continuous replication.

If i put my "critical" VMs on a separate storage (maybe NSS 322) to enable the great vMotion capabilities, it is important for me to have a redundancy and an automatic failover. A network failure (MPIO) would be nice, but not 1st priority. In a secondary step i am looking for the possibilities with link aggregation and MPIO...

Syncing two NSS via rsyc or 'dd' maybe would be a simple solution, but these functions are working as scheduled task and do not permit an immediate failover. Besides it is not possible to rsync incrementally, so the hole vmdk or image files have to be copied. Some of my vmdk files are greater than 500 GB!

After many hours of searching i found different interesting products: Open-E DSS, Veeam, Starwind, SanMelody..etc. but these solutions seem not to fit with 2 NSS 322. A storage virtualization solution like FalconStore would be the best for me, but FalconStore is too expensive.

So, im still searching for a (cheap) solution...

Many thanks for your help!

Greetings from Switzerland