Hi Kevin

First, I think your current config is unsupported, since max FCP nodes in a single cluster is 8 in cDOT 8.2.  (Going from memory, but I’m pretty sure that’s the rule.)

Where did you read that zoning guideline?  I’d be surprised if that was a NetApp practice.  If a switch or server/OS vendor, then it merits investigation.  If another storage vendor, then I’d say that’s their problem and you should feel confident in the NetApp practice of a zone for each initiator to target LIFs on each node x fabric.

Another tidbit for you is that in 8.3 we have a feature called Selective LUN Mapping.  When new LUN maps are created, by default the only paths visible will be on the node owning the aggr containing the LUN and its HA partner.  This will automatically reduce visible paths.  Note that the main value of SLM is to reduce path consumption in attached hosts/servers.  For example, VMware ESXi allows 256 LUNs and 1024 paths (including non-array devices such as USB, CD, internal disks, etc.).  With more than two cDOT nodes, you start to deplete paths faster than LUNs.  Having said that, depending on the issue that 4 target practice is meant to solve/prevent, SLM may do the job in another, simpler way.

Share and enjoy!

Peter

From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Kevin Richards
Sent: Monday, July 13, 2015 1:53 PM
To: toasters@teaparty.net
Subject: fcp paths in multi-node cluster

Hi,

I have read that an initiator should not be zoned to more than 4 target ports.  We have a 8.2 cdot cluster with 12 nodes each having two fcp lifs.  So there are potentially 12 paths to storage per fabric.

Just curious what strategy has been used to limit the paths?  Zoning, port sets, etc.? And any recommendations.

Thank you.

Kevin