Multi site question

[NzO]

Hello,

Forgiveness in advance for my English.

I have several question about managing multiple sites.

Even if a site is declared as primary, a LUN can be active on another site. Is there a way to manually switch an active node to another?

Why multiple LUNs use the same LeftHand as gateway connection while others are available? Can you define it ?

In my case I have 6 nodes, 4 LUN of 2 To,  2 LUN use my Lefthand1 as gateway connexion.

The load is not balanced and I can have large peak of performance on my switch.

In the capture image001, the port 1 & 5 is my lefthand 1 and the port 22 & 24 my trunk to the other switch /site

My configuration :

3 p4300 per site configure in ALB mode

1 switch 2910 procurve per site with a 2go trunk

2 esx 5 per site 

i follow this best practice : Best practice

[Amar_Joshi]

NzO,

The document you have referred for best practices from VMware is the correct one and I am hoping that you have gone through page8 - Multipathing iSCSI and page9 - Enabling LB for performance. These are the 2 key aspects of achieving the load balancing across multiple nodes.

Some checks are needed for the configuration sanity check -

- Do you see multiple paths on ESX under Storage Adapters > iSCSI software initiator > LUN > Manage Path

- From the CMC validate you have multisite cluster

- From the CMC make sure you have sites assigned and each has 3 nodes in it

- From the CMC check if clusterwide load balancing (VIPLB) is enabled or not?

Q: are you using single network subnet for all the nodes or they are in 2 different subnets (one for primary and other for remote site)?

Here is the older best practices guide (for ESX4) which has more better screenshots for a better understanding of multipathing setup. Please refer to pg6.

http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA3-0261ENW.pdf

Hope that helps and please let us know your resolution.

[NzO]

I follow the best pratice for the iscsi multi path too

I see multi path in storage adapter, 

I have a multisite cluster 

I have got 2 sites with 3 nodes per sites and a third site for the FOM.

I have 1 vip address actually managed by the Lefthand2 (lefthand host on the 2nd site)

And the 6 nodes are on the same subnet.

My datastore are configured with round robin mode and recently i configured the IOPS=1

[Jitun]

Hi NzO,

From the screenshot, somehing is not right.

All the volumes have Node1 as the gateway.

Please right click on servers in CMC and select Edit.

You shall find an Checkbox for Enable Load Balancing.

I hope they are checked for all servers.

Could you please mention how the servers are assigned sites ?

Interesting read

http://www.virtuallifestyle.nl/2011/11/hp-p4000-multi-site-san-and-vmware-vsphere-watch-out/

[NzO]

Hi,

Thanks for the link.

I agree, something is not right, but load balancing is check on all servers.

I have two other volume that uses lh2 and LH3.

Thing I do not understand either why this is not LH5 from site 1 that is not used ?

even number site 1 

uneven number site 2

The FOM is on a third site unassigned

Tomorrow I will try the plugin vmware vcenter for storage array 

[Jitun]

Please confirm your MultiSite is Spread across 2 Subnets ? You have 2 VIPs ?

If yes, Please mentioned if ESX01 and ESX03 have one of the Subnets as the primary

and ESX02 and ESX04 have the 2nd Subnet.

If not, if you have a Stretched Subnet (single VIP), then if you can, could you try disconnecting a Volume assigned to Site2 and see if the Gateway changes to one of Site2 P4000 Nodes..

Also try stopping the manger, on the Node which is the Coordinating Manager(you can check this by clicking on Management Group in CMC) and starting it back ON. This will another manager to become the Coordinating manager.

[NzO · Accepted solution]

Multisite is configure with only 1 subnet and one VIP adress.

I removed the ESX servers in the sites, they now unassigned.

After that, I stop / start the manager on LH1 and the gateway connexion are balanced now.

Now, each node that has the manager started is used as a gateway connection by different LUN.

Site management for ESXi is not great

[Jitun]

Glad to know the issue is Resolved.