Can I resolve queue depth issue by moving FC links to additional HBA?
Applies to
- ONTAP
- Fibre Channel
Answer
- FC queue depth is per port not per Host Bus Adapter (HBA) so this approach would not resolve the underlying port issue of SCSI queue exhaustion.
- You need to reduce the initiator queue or increase the target queue depth by adding more ports and moving some paths to new ports.
Example:
- Each storage HBA port could have a queue depth available of 1945 ( 2048 -5 % for critical SCSI commands ) moving FC link to another HBA port would not change this.
- This means 32 hosts initiator ports with a queue depth of 64 can access the target port
Additional Information
NOTE: If receiving an error combined with a performance impact Performance issue with STIO TPD cmd alloc threshold reached handle
Alternatively :
- Consider tune the environment and Hosts Calculate queue depth overview.
- Confirm in Hardware Universe supported number of LIF's per port for your platform.
- Hardware options and supported HBA can also be found in Hardware Universe.
- Consider add additional target ports and FC links
- Example: from 2 target ports per node go to 4 this may alleviate the queue depth pressure on each port.
- Consider moving the environment to a Storage system / HBA card with increased queue depth support for example 4096 ( - 5 % for critical SCSI commands).
- Alternatively lower the Host queue depth to meet limits
To Check queue depth per port:
network fcp adapter show -fields queue-depth