Need to remove DAE on CX4-480 flare code 4.30.x

USM doesn’t allow you to remove DAEs….only add them.

Joe Sullivan

Storage Administrator, Enterprise Storage

Information Services

Indiana University Health

South Campus, 1515 N Senate Ave., Indianapolis, IN 46202

317.962.5415

jsullivan4@iuhealth.org

Discover the strength at www.iuhealth.org

You can download the CLARiiON Procedure Generator from Powerlink which would give you step by step instructions.

Moreover, I reckon in order to remove a blank DAE you can run Install/Uninstall Hardware wizard from USM and that would be it!

Technically this is possible, I don’t think the procedure is documented and you will want to make sure EMC support is engaged.

Richard J Anderson

Richard Anderson wrote: IF… 1.)    The DAE is at the end of the loop Richard J Anderson

isn't there a procedure to remove DAEs that are not at the end of the loop, one LCC at a time ?

Oh Sorry, my bad. I just checked CPG, there is a procedure there well you can follow that if you like but basically its just I initially thought Just remove it!

Then restart the management server to remove the ghost entries in Navisphere/Unisphere GUI.

All done and working great!!!

Thanks for your input!

Happy T-day

Joe Sullivan

Storage Administrator, Enterprise Storage

Information Services

Indiana University Health

South Campus, 1515 N Senate Ave., Indianapolis, IN 46202

317.962.5415

jsullivan4@iuhealth.org

Discover the strength at www.iuhealth.org

I have a customer that needed to do it on a VNX and at first support said it couldn’t be done, then they later “created the procedure and tested it in the lab” so my impression is that it’s not in even the internal CPG, nor is it in any public documentation. So that’s why I suggest getting with support if you actually need to remove a DAE in the middle of a loop.

Richard J Anderson

Richard Anderson wrote: Technically this is possible, I don’t think the procedure is documented and you will want to make sure EMC support is engaged. Richard J Anderson

that procedure is not documented for customers or for EMC support ?

The procedure is not mentioned seperately in CPG. However, its there when you run Uninstall a DAE/DAE2-ATA/DAE2P/DAE3P enclosure from an active storage system CPG.

Only thing you need to keep in mind is when you remove the HSSDC Cable from the LCC A of the Encl you want to remove you would need to plug the lower encl's Cable to the one above it and follow the same for LCC B.

I will post a detailed post with a diagram soon. (Sun or Monday - gtg)

yes

dynamox wrote: isn't there a procedure to remove DAEs that are not at the end of the loop, one LCC at a time ?

You mean DAE not the top DAE but any other DAE on the array except Vault DAE?

nice, maybe you can ask glen to turn it into a document ?

I have access to that and I have created several Primus Articles myself but yes, Im Okay if he wants to do it!

Removing a DAE from the middle of the bus for a VNX is not possible.  It has to do with the way the system automatically numbers the DAE's. 

emc299019: "Can a DAE in the middle of a VNX backend bus be removed?"

[...]

The removal of the a DAE in the middle position of a backend bus should only be used for the replacement of that DAE. This action will cause a data unavailability situation for the enclosure to be replaced, and any subsequent enclosures with higher numbered ID's connected on the same backend bus. The DAE cannot be removed permanently.

[...]

However, for previous generations not using the automatic numbering, the process would go as follows:

PREPARATION:

============

1) As mentioned already, there is nothing requiring that the DAE numbering is contiguous on the back-end buses (except for the VNX).

2) It goes without saying that make *absolutely* certain that every drive from the trays being removed are unbound.  Look also for private LUNs.  Cleanup may be messy if you leave anything orphaned.

3) Unless you plan to rerack during the process, I recommend ahead of time you verify that the cables reach between DAE's

EXAMPLE (removing 2 DAE's from middle of bus):

========

For this example, I'm going to assume you are looking to remove trays 0_1 and 0_2 from the backend bus currently hosting:

0_0, 0_1, 0_2, and 0_3

Therefore, when completed, only 0_0 and 0_3 are left behind.

SPA

====
1) If the array is running FLARE 26, it supports the lower-redirector (introduced with ALUA but technically isn’t ALUA) so it will not trespass LUNs immediately as it will redirect back-end I/O via its peer. 
- You don’t (necessarily) need to trespass LUNs as it will utilize a non-optimal path in the backend (hop through its peer)

2) If the array is running < FLARE 26 or they own the single-ported SATA drives (very old), trespass all LUNS from tray 0_3 first to SPB (so the system doesn’t have to on its own)

- Verify first that the SP’s aren’t over 50% utilized (threshold is 60%) prior to trespassing all LUNs

3) On DAE 0_3/LCC A/port “Pri”, disconnect cable (will be cable coming from 0_2/LCC A port “Exp”)

4) On DAE 0_1/LCC A/port “Pri”, disconnect cable (will be cable coming from 0_0/LCC A port “Exp”)

5) Connect cable from DAE 0_0/LCC A/port “Exp” to DAE 0_3/LCC A/port “Pri”

** Before you continue, make sure that DAE 0_0 and 0_3 are not faulted (0_1 and 0_2 will be) or rather that Navisphere (or Unisphere) acknowledges the redundant paths

SPB
===

6) If the array is running FLARE 26 or greater, it supports the lower-redirector (introduced with ALUA but technically isn’t ALUA) so it will not trespass LUNs immediately as it will redirect back-end I/O via its peer. 
- You don’t (necessarily) need to trespass LUNs as it will utilize a non-optimal path in the backend (hop through its peer)

7) If the array is running - Verify first that the SP’s aren’t over 50% utilized (threshold is 60%) prior to trespassing all LUNs

8) On DAE 0_3/LCC B/port “Pri”, disconnect cable (will be cable coming from 0_2/LCC B port “Exp”)

9) On DAE 0_1/LCC B/port “Pri”, disconnect cable (will be cable coming from 0_0/LCC B port “Exp”)

10) Connect cable from DAE 0_0/LCC B/port “Exp” to DAE 0_3/LCC B/port “Pri”

** At this point you will likely see 0_1 and 0_2.  I’ve been able to clear those in the past by restarting the mgmt service on each SP.  I’ve heard though of people having to wait until they are able to fully reboot each SP one at a time; however, I believe it is because something was left behind and the drives weren’t fully unbound.