Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

4271

March 28th, 2016 11:00

Change Initiator type XtremIO

Hello,

We have XtremIO hosting all ESXi hosts and the initiator type is configured as "other" instead of "ESX". Is it safe to change the initiator type to "ESX" when the host is online or doe we have to unmount the LUNs, take the host offline and then change the initiator type?

Thanks

64 Posts

March 29th, 2016 07:00

This setting can be changed on the fly with no impact to the host.

This setting (currently) only controls the behavior of the array during things like upgrades and outages - it doesn't affect anything related to the normal operation of the array. What's more, as of version 4.0.2 both the "ESX" and "Other" settings result in the same behavior - although that will likely change in a future version as we tweak it a little for ESX.

So yes, you can change it any time, without any need to do anything on the host. The host will not even detect the change in any way.

522 Posts

March 28th, 2016 12:00

It should be done offline since it can change the SCSI personality exposed to the host as the best practice (I equate it to FA bit settings, or Clarion mode settings). However, I just did this with a customer (non-prod though) where we changed it one initiator at a time with a rescan and it worked fine.

What you cold also do if you want to do it a host at a time is to drop it into maintenance mode and do it that way.

522 Posts

March 29th, 2016 07:00

Thanks Scott for the clarification....that's interesting since support/EMC had told me different no too long ago...I might suggest adding this to the list of updates to the documentation if it hasn't been added there already. A lot of people, without this type of documentation, might equate it to the FA bits and VNX settings, which are not 100% online changes for all values.

Also, can you provide some documentation or explanation of what these settings do with regards to the upgrades and outages? I haven't seen any documentation on those impact areas. Thanks.

-Keith

727 Posts

March 29th, 2016 13:00

Hi Keith,

I have reached out to the documentation team and we will be adding this information in the user guide and host configuration guide.

Thanks!

522 Posts

March 30th, 2016 05:00

Thanks Avi.

Do you or Scott have any info on the other question prior to the documentation updates about how these initiator types are leveraged with regards to upgrade, outages, etc? What is the ramifications of them not being set and what does setting them buy us during an upgrade? I am not aware of the recent health checks done for upgrades to call these out or validate them so it seems to be more like a nice-to-have versus a requirement?

Thanks!

-Keith

727 Posts

March 30th, 2016 19:00

The Host Configuration Guide specifically calls out that this change should be done for Solaris environments. In those cases, we do change the way XtremIO interacts with the Solaris system – and this was done to conform with the expectations from the host side.

Note that we may make changes for other OS types also in future and it may be done to ensure stable and optimal performance with those operating systems. I will try to find more details and share it here.

727 Posts

March 30th, 2016 20:00

See the KB article for Solaris recommendation at https://support.emc.com/kb/204507.

522 Posts

March 31st, 2016 05:00

Thanks Avi.....I am good with the Solaris changes - those I know about and those are called out upon customer upgrades. I was more curious on the other OS's.....they have drop downs, but I don't know of any documentation/KBs/etc with relation to what those do during the upgrades and outages that scotthoward wrote about, so I was hoping to get a more generic understanding of what they do. For example, if a customer doesn't set them or migrates from 3.x where they aren't set....it seems they can flip them online (which is good)....but is it required? I don't see them called out in the upgrade process or documented like the Solaris settings you have mentioned, so are they just nice to have? Best Practice? useless for host interaction? That is the stuff I am trying to qualify with Scott's response.

Typically I have educated customers to move to them sooner than later as a best practice, but again that was based off the information I got from support around host presentation and that made me infer them as being more "required" than nice to have (Think of FA bit settings or VNX settings for host registration). But if the setting doesn't control or touch any of that, I'd like to understand more what it does touch and where it is required.

Thanks!

-Keith

64 Posts

March 31st, 2016 08:00

As at today, none of the OS settings does anything specific other than "Solaris" - all of the others acts the same, which is the same as "Other".


However if and when we find optimizations that we can make for different operating systems then this may change, and if you're not set to the right OS then at best you won't get the benefits of those changes (and at worst, if you're set to the wrong OS, then something might break due to the wrong behavior for that system).

As an example, we're currently in testing for changes to the "ESX" setting which will likely go into a future version.  If your ESX servers are set to "other" then you won't see that change.

So in general it's best to set them to the correct values, but (with the exception of Solaris) if they are wrong today it's certainly NOT a critical issue that needs to be addressed immediately - but it's definitely worth fixing it at some stage.

522 Posts

March 31st, 2016 09:00

Thanks Scott....so just to clarify, other than Solaris, the settings don't do anything today - even for upgrades and outages it seems - I just want to make sure there is no difference in host presentation and outages/upgrades today by setting these values given the discussion on this thread mentioned above.

Thanks!

727 Posts

April 1st, 2016 14:00

That’s correct Keith.

No Events found!

Top