Upgrading to SVC V5.1.0.x from V4.3.1.x can lead to single nodes repeatedly rebooting, which in turn can lead to the upgrade process stalling, if the cluster has any Incremental FlashCopy mappings defined.
During a cluster upgrade, host I/O to a VDisk in an Incremental FlashCopy mapping involves messages being passed between the IO Group(s) containing the source and target VDisks, and the IO Group containing the bitmap information for the mapping.
If a SVC node that is running V4.3.1.x queries a node that has been upgraded to 5.1.0.x for information on the Incremental FlashCopy mapping, there is a possibility of it receiving back invalid data and consequently asserting (rebooting) repeatedly. This node will then be excluded from the cluster and the upgrade will enter the Stalled state. Depending on where in the upgrade process this condition was encountered, the stalled upgrade may need to be aborted and rolled back, which may lead to a loss of access to VDisks in the affected node's IO Group.
This issue was resolved by APAR IC70105 in the V188.8.131.52 PTF release.
If customers with Incremental FlashCopy mappings have a requirement to upgrade to a V5.1.0.x level earlier than V184.108.40.206, please either delete all Incremental FlashCopy mappings prior to commencing the upgrade, or alternatively contact IBM Support for assistance in determining whether this cluster's configuration is exposed to this issue.
Internal Use Only
21/04/2010 - Wording provided by Ed Moffatt, reviewed by Steve Whilte. CMVC 111435
23/09/2010: EH/ GL reviewed. This document relates to SVC V5.1.x and earlier only.
20/10/2010: SW updated to remove statement of this being fixed in 220.127.116.11 via APAR IC65667, and to state that this will be fixed in a future release via APAR IC70105.
7/12/10: SW - Fix available in 18.104.22.168
17 June 2018