Example jira DellEMC PowerPath / EMC Solutions Enabler / EMC Cluster Enabler Interfering with PURE LUNs
Symptom:
- cluster.log shows PR reservation issues with Pure LUN such as (note the PR key a72a40d00001734d used by the cluster)
INFO [RES] Physical Disk <Cluster Disk 1>: Arbitrate - Node using PR key a72a40d00001734d
ERR [RES] Physical Disk <Cluster Disk 1>: OnlineThread: Unable to arbitrate for the disk. Error: 2.
Recommended actions:
Please reach out to Everpure support to carry out additional troubleshooting on the storage end and provide further recommendations.
- If customer is hitting SITUATION 2, please first ensure that the LUN PR reservation was done with the PR key found in the cluster.log (note - only the line with the Key is displayed for brevity):
root@array-x-ct0:/var/home/puresupport# date; puredb list reservation Fri 05 May 2023 08:57:44 PM AEST Vol Fabric Initiator Target Isid Reservation Key Type 75925 fc 10000090FA8C5FAE 524A93739733F601 0 0xa72a40d00001734d 5 flash_array_LUN - Ensure that no sense reason 16 hits can be found on neither controller on the storage array. If you do find hits for this, it means there is an actual PR conflict which would have to be investigated deeper:
rkukurs@cloudfuse-prod-02658d463d14bedec:/logs/domain.au/array-x-ct0/2023_05_05$ zgrep -c 'sense_reason 16' kern.log
Conclusions:
If there are no PR conflicts (sense reason 16), and the PR has been placed on the LUN in question, the most likely cause for the cluster being unable to arbitrate is DellEMC software. Please ask customer to verify if they use EMC Solutions Enabler / EMC Cluster Enabler to manage their DellEMC cluster LUNs. If they do, that is the most likely explanation - those services do not allow the cluster to use a valid key to register against Pure LUNs.
Sample POA to propose to the customer FOR SITUATION 2 if their goal is to migrate data from DellEMC to Pure (it was done in DellEMC PowerPath / EMC Solutions Enabler / EMC Cluster Enabler Interfering with PURE LUNs):
- Shut down all the cluster resources but leave the disk resources up and running
- Mount Pure LUNs to OS (not to the cluster yet) and copy over data from ENC LUNs
- Remove EMC LUNs from cluster roles so that just the SQL and network resources are left
- Open the "EMC Cluster Enabler Manager" and "de-configure" the cluster (this should not impact the SQL resources left in the cluster, this was required to be able to uninstall the EMC software or the installer says you can't uninstall with configured cluster)
- Start SQL resources with storage mounted locally outside cluster with the correct drive letters
- Create required csv LUNs on Flash Array / Pure end
- Add csv LUNs to the cluster without errors if everything has been done properly