topic Re: Lefthand cluster best practices in StoreVirtual Storage

Lefthand cluster best practices

kins — Tue, 24 Nov 2009 17:48:45 GMT

Hi,

Is it possible to place one of the two nodes (4500) in a remote DR site and replicate between the nodes?

Is a licence required for replication over different subnets?

With a 2 node cluster it is recommended to install the failover manager away from the SAN for obvious reasons. If you do not have a third site would you install it onto a physical Windows server or create a Windows VM and store on local disks.

Re: Lefthand cluster best practices

René Loser — Wed, 25 Nov 2009 09:41:07 GMT

Hi,

Latency must be less than 2ms (roundtrip).
No license is required. Everything is included. So you can splitt the 2 nodes over 2 sites.

Best practices is to have FOM (failover manager) on a 3rd location. At least on an independet machine running. FOM runs on VMware Player/Server/ESX.

regards,
reNe

Re: Lefthand cluster best practices

kins — Wed, 25 Nov 2009 10:08:12 GMT

thanks

Re: Lefthand cluster best practices

kins — Wed, 25 Nov 2009 10:09:56 GMT

What sort of latency is recommended for remote copy?

can remote copy work with any storage as a target?

Arfe there any links for RC?

Re: Lefthand cluster best practices

teledata — Wed, 25 Nov 2009 15:46:50 GMT

>>What sort of latency is recommended for >>remote copy?

There are not latency requirements for Remote copy snapshots since they are asynchronous. There are bandwidth requirements however. And it varies based on how much data (and how frequently) you snapshot. A good rule of thumb is that a single T1 (1.54Mbps) can replicate about 15GB of data per 24 hours. That is fully saturated, so make sure you won't be competing with other WAN traffic.

>>can remote copy work with any storage as a >>target?

It must be Lefthand SAN/iQ, so that would be another P4xxx SAN module or the P4000 VSA (which can run on any storage that can run as an ESX(i) host). If you want to use the VSA for managed snapshots it must be licensed, or acquired through the included bundle (with the virtualization SAN product)

It sounds like you do NOT have a low latency connection between locations (right?)
If that is the case, and you want to be able to sustain a single storage node failure (and keep critical storage online) at the primary site, you will need to have the 2 SAN modules at 1 location and configure critical volumes with 2-way replication. If in addition to that you want to have off-site snapshots (using remote copy) you will need to purchase a 3rd SAN module, or purchase the VSA license to run on Bring-Your-Own-Storage (this can be run on the free version of ESXi 3 or 4).

Re: Lefthand cluster best practices

kins — Thu, 26 Nov 2009 08:51:17 GMT

Hi thanks for replying.

Yes we have a 10MB link between the sites. We have a 2 node cluster in the HQ and a VSA for the DR site. ESXi is running on the VSA box and will be moved shortly.

Does Primesync work with the VSA so we can replicate data locally before moving?

In this scenario with 2 sites where would you place the FOM? I know I can't put it on the 2 node cluster. The way I see it I have 2 options:
1. put it on local disks on one of the ESX hosts
2. put it on the VSA in the remote DR site

What do you think?

Thanks again

Re: Lefthand cluster best practices

kghammond — Tue, 01 Dec 2009 13:41:52 GMT

In regards to the 2ms latency for a multi-site cluster, does anyone know if this is a hard requirement?

We have been designing a LeftHand envionrment that will span multiple sites on a 10G fiber, but I am not positive that we will be able to gurantee 2ms latency.

Thank you,
Kevin

Re: Lefthand cluster best practices

René Loser — Wed, 02 Dec 2009 06:54:26 GMT

Hi,

The <2ms is the recommendation in average for a Multi-Site Cluster.
The tolerance is around 5ms max.

Best regards,
reNe

Re: Lefthand cluster best practices

VMSE — Wed, 13 Jan 2010 18:25:32 GMT

Is FOM (failover manager)same as SANiq management console that shipped with the SAN? I often see different names tossed around but not sure what they are.

Thanks.

Re: Lefthand cluster best practices

kins — Thu, 14 Jan 2010 17:21:12 GMT

Hi the MGMT console is called "Central Management Console".

FOM is "FailOver Manager" and is required to maintain Quorum in the event of a node failure in a 2 node cluster.

Both ship with the SAN

Re: Lefthand cluster best practices

teledata — Fri, 15 Jan 2010 17:46:10 GMT

I always try to install the FOM on local storage. (so it is not dependant on the same storage it is protecting).

The other benefit, is in the case of a reboot of the ESX host that FOM is running on, local storage is always going to be available (assuming you've configured the FOM VM to autostart). You can have the case where if you have a power failure, then power returns, it is possible that your ESX host(s) may finish booting before your SAN modules do. I always play it conservative with the FOM because I don't want anything impeding it from automatically launching as quickly as possible in the event of some kind of failure.

Re: Lefthand cluster best practices

Chris Davin — Tue, 16 Mar 2010 14:48:57 GMT

Hi

I have a designated 10mb connection for our DR site but grinds to a halt at certain times of the day unless we increase the available bandwidth, the latency is normally 1ms so technically this should not be a problem. Is there an option to buffer data for low bandwidth connections (10mb)?

Thanks