Hello Andy,
besides what Nezih wrote some additional things. On the SVC you have to check if backend unmap is turned on, in more recent firmwares of Spectrum Virtualize this is usually turned on automatically.
Additionally on the FCM storage you have to check if hostunmap is enabled, this is not the case by default. On HDD storage this should be turned off.
The commands for this are:
SVC: "chsystem -backendunmap on"
Backend FCM Storage: "chsystem -hostunmap on"
Backend HDD Storage: "chsystem -hostunmap off"
Check as Nezih wrote.
But: Storage which is currently not released, because the backend FCM controller does not know about it, will not be released automatically after the change.
At least that is what we have found, for this the storage must be released again on the host systems and/or SVC.
E.g. with the tool sdelete in case of windows LUNs
For this the host and svc host configurations must be checked if they can use unmap and undelete. Otherwise the SVC will not get any information if memory is freed or not in a connected system.
Also keep in mind that the memory systems should be kept on a current level for stability and security reasons:
https://www.ibm.com/support/pages/spectrum-virtualize-family-products-upgrade-planningGreetings Patrik
Source:
https://www.ibm.com/docs/en/sanvolumecontroller/8.3.x?topic=commands-chsystem------------------------------
Patrik Groß
------------------------------
Original Message:
Sent: Sat September 03, 2022 06:15 AM
From: Nezih Boyacioglu
Subject: FCM behind SVC releasing capacity?
you need to check backend unmap settings on SVC and backend FCM storage ( lssystem | grep unmap )
you will find more information on Performance and Best Practices Guide for IBM Spectrum Virtualize 8.5 Redbook starting in Chapter 3.3.2 https://www.redbooks.ibm.com/abstracts/sg248521.html
------------------------------
Nezih Boyacioglu
Original Message:
Sent: Fri September 02, 2022 08:57 AM
From: Andy 91717
Subject: FCM behind SVC releasing capacity?
We've gone through a set of actions in one of our test environments whereby we can't make sense of what the capacity reporting that the SVC and backend FCM arrays are showing. To describe our environment:
- Front end SVC cluster, running 8.3.1.3 code
- Backend FCM storage running 8.4.0.4 code
- Backend HDD storage running 8.3.1.3 code
On the SVC I did the following:
- Moved LUN's from dedicated FCM mdiskgrp to dedicated HDD mdiskgrp
- When I checked the mdiskgrp on the SVC, it went down the balance of the LUN's moved, as expected
- But when I check the backend FCM storage array there was no change in usable capacity stats for it's storage pool
The LUN's that moved back to disk storage pool, according to SVC, have a very low compression rate, so I would have expected the FCM storage usable capacity to go down also. Is there some sort of cleanup that needs to occur on the FCM array to allow this data which has moved to be reclaimed, or something like that?
------------------------------
Andy 91717
------------------------------