VIP's in Multisite cluster, explanation

[Anonymous]

Guys,
can someone summarize me what will happen to file server in multisite configuration that is configured with two VIP’s (one per subnet).

So for instance we have:
Site A in subnet 192.168.1.0/24, VIP-A address 192.168.1.10 with cluster storage units S1 192.168.1.20 and S3 192.168.1.21
Site B in subnet 192.168.2.0/24, VIP-B address 192.168.2.10 with cluster storage units S2 192.168.2.20 and S4 192.168.2.21
We have FA agent installed so it will see if one of the sites fails.

So we have a server which is in 192.168.1.0 (site A) subnet with ip address .100. iSCSI target for server is configured to 192.168.1.10 and file share volume is discovered (VOL_SHARE). iSCSI gateway is chosen depending on the load balancing when server is started (it will be either .1.20 or .1.21). VOL_SHARE is in Network Raid 10.  And now SAN Site A fails completely, due to hardware accident. File server was online and connected to gateway while this happened. Site B continues to function without a problem.

QUESTION:
What will happen to file server that expects iSCSI respond from 192.168.1.10 (actually from iSCSI gateway) that is now down? How should server be configured to contiune to run normaly if site failes?

[Emilo]

The key to this question is in the proper setup of the sites in the Multi-San site. Its the sites and the VIP-LB (virtual IP Load balancing. I/O path preferencing based on site/subnet so the application you describe will automatcly load balance and select a new gw. VIP-LB simply terminates the iSCSI session on local Storage Systems. With the proper setup you should have the data replicated on the B side. Make sure you use the wizard to setup.

[Anonymous]

Emilio, than you. Where should I/O path referencing be configured? In Windows OS in this case or on SAN?

This question is hypothetical, situation does not exist. We are thinking on buying multisite but this scenario is never explained. Can you give me maybe some pdf where this is explained. I do understand what will happen if I have only one subnet, but I do not understand in two subnet scenario. Because you can't move VIP from subnet A and just "host" it on B. Router will not route traffic to it, it's simple as that.

Can you possibly explain, if everything is properly configured what will happen with this scenario if one unit fails on site A and both units fails on site A?

[Anonymous]

I found it, "Failover   No load balancing is performed. The application specifies a primary path and a set of standby paths. The primary path is used for processing device requests. If the primary path fails, one of the standby paths is used. Standby paths must be listed in decreasing order of preference (the most preferred path first)."

So actually if Windows client connected on Site A detects path failure it will just switch to secondary path. So clinets are the on that will do failover, and SAN is responsible only to VOL synchronization. 

[oikjn]

IF you are on the same subnet (such as myself), I used the HP DSM provider to handle the local/remote node connections. 

My setup:

five VSA nodes, three in one server room and two in another on the other side of the building.  Rooms where connected through a single trunked pair of switches.  My FOM is in the same room as the three nodes in the main server room.  I have a single subnet for all iscsi traffic. and all five VSAs are in the same cluster.

I assigned the FOM to site "sdfsduhkf" (name doesn't matter), then I assigned the free VSAs in the main room to site "server room".  I also assigned all my servers in that room to the same site.  I then assigned the two remaining VSAs to site "backup site".

I connected the iSCSI connections to each lun as instructed using the HP DSM guide and the result is that the servers only connect to the VSAs on the same site as they are assigned (in a failover situlation they automatically connect to the remote site VSAs).

I don't know if there is that option for VMWare, but in windows with the HP DSM, this is pretty smooth and easy.  Just search for and read the documents on HP DSM.

[Anonymous]

As far as I know ther is no multipathing plugin for vSphere. Per volume only one gateway not like in Windows all nods in cluster (due to MPIO).

[oikjn]

I think you are correct about the vmware limitation.  I really only deal with windows iscsi and their MPIO/DSM features work great.

I can't comment about what to do for VMWare, but I"m sure there has to be something like MPIO functionality

[Bart_Heungens]

Hi,

There is currently only a DSM available for Windows, not for VMware. It is already a long time demanded/awaited but untill further notice it isn't there yet...

However, what you can do to obtain the best performance with VMware and Lefthand is enabling load balancing on all paths.

All details in the following document

http://h20195.www2.hp.com/V2/GetPDF.aspx/4AA3-6918ENW.pdf

Can tell you that it works fine, easy to setup (in v5, v4 requires some command line actions) and gets most out of your storage nodes...

Kr,

Bart