Troubleshooting disk foreign state issue when replace a disk on DCA

Products

• VMware Tanzu Greenplum

Issue/Introduction

Symptoms:
When replacing a disk on DCA V1 or  DCA V2, the replacement disk shows up as foreign state if the replacement disk is from another system.

Environment

Cause

When a controller firmware detects a physical disk with existing foreign metadata, it flags the physical disk as foreign and generates an alert indicating that a foreign disk was detected.

Resolution

Verify disk foreign state

DCA V2

# CmdTool2 -pdlist -aall | egrep "Adapter|Enclosure|Slot Number|Firmware state”
Slot Number: 6
Enclosure position: 0
Firmware state: Online, Spun Up
Enclosure Device ID: 8
Slot Number: 7
Enclosure position: 0
Firmware state: Online, Spun Up
Enclosure Device ID: 8
Slot Number: 8
Enclosure position: 0
Firmware state: Unconfigured(good), Spun down
...

DCA V1

# omreport storage pdisk controller=0
ID                       : 0:0:10
Status                   : Non-Critical
Name                     : Physical Disk 0:0:10
....

You can clear the foreign state on the RAID controller.

DCA V2

CmdTool2 -cfgforeign -clear -a0

DCA V1

omconfig storage controller action=clearforeignconfig controller=0
omconfig storage pdisk action=assignglobalhotspare controller=xx pdisk=xx:xx:xx assign=yes

 Note

• Clearing foreign state on the RAID controller is different than importing foreign configuration, which happens when you replace the entire server but put the original disks back.

• After clearing disk foreign state, the disk rebuild won’t be kicked out automatically. For DCA V2,  firmware state of the disk shows “Unconfigured(good)”. You need to manually run the disk rebuild. Please see DCA V2 maintenance guide for detailed steps. For DCA V1, we haven't seen an example of such problem yet and we don't have valid commands to force the rebuild. If you notice DCA V1 disk didn't start rebuilding, you may try reseat the disk after clearing the foreign state. If that doesn't work, dispatch a new disk for replacement. Open a Jira to track this issue.