in Selective Lun Map, what if the HA pair owning the lun got failed?

[netappmagic]

>With SLM, when a new LUN map is created, the LUN is accessible only through paths on the node owning the LUN and its HA partner.

In a 4 nodes cluster (2 HA Pairs), what will it happen when HA pair owning the lun got all failed? Clients then could not access the lun at all, not even through indirect access, correct?

Please share your ideas. Thanks!

[SeanHatfield]

In that scenario the LUN wouldn't be accessible anyway because the hosting aggregate would be offline.

[heightsnj]

A follow-up question.

in the same 4-nodes cluster, what if "reporting-nodes" reports all 4 nodes, would that have any impact on performance?

Since normally "reporting-nodes" should only reports 2-nodes in HA owning the lun, not all 4 nodes.

[SeanHatfield]

That depends.  The primary reason for limiting the reporting nodes is to help avoid path depletion.   For example, a VMware host can support a theoretical maximum of 1024 SAN paths.  So the maximum number of luns you can support is a function of the number of reporting paths to each lun.  If each lun has 16 available paths you could support at most 64 luns to a given VMware host. If each lun had only 4 available paths, that number could be 256.  So Selective Lun Mapping, in conjuction with portsets, becomes increasingly important in a scale out SAN use case as the node count increases within the cluster.

But if you are not approaching the lun path limits of your SAN attached hosts, and the host multipathing software is functioning correctly, there is no harm in having the additional reporting nodes.  

[heightsnj]

Based on what you said, reported not-owning HA nodes by "reporting nodes" should not cause any performance issues, since SLM or reporting nodes only for the purpose of reducing the number of LUN paths.

Then the next step I think we should look into is if the host is using the optimized the path. If mulitpathing is configured, and ALUA is enabled, what will be the way or commands to verify if the optimized path is being used by the host access on AIX/Linux server? assuming there is no way to verify it on the filer.

Thanks!