High write latency in Azure CVO
- Views:
- 507
- Visibility:
- Public
- Votes:
- 0
- Category:
- cloud-volumes-ontap-cvo
- Specialty:
- cloud_performance
- Last Updated:
Applies to
- ONTAP 9
- Azure
- CVO (Cloud Volume ONTAP)
Issue
High write latency in Azure CVO (Cloud Volume ONTAP) caused by slow NVRAM transfer when using page-blob disks
Cause
- The communication with the page-blob disks is done using API calls over HTTP/S
- If that network is slow, writes to disks will be delayed and will result in write latency
Solution
- Check if the increased workload is expected and can be reduced.
Contact Microsoft to address the Azure network and the Azure disk layers and validate resource contention.
Additional Information
additionalInformation_text
Internal Notes
- See burt 1533129
- The issues with the high write latency were observed on multiple layers of the CVO system, indicating an issue with the backend disk used by the system
- The workload view shows high WAFL Suspend CP latency, indicating issues while trying to write the data to the disk:

- This can be confirmed from the Consistency Point (CP) view, showing Back-to-Back (B2B) CP events:
.png?revision=2)
- These events had higher P2_FLUSH latency, confirming that the writes to disk on Azure were the root cause:
.png?revision=2)
- We can also observe NVLOG_FULL Messages for SPINNP_WRITE in WAFL Suspend:
.png?revision=2)
