Incident Report: CVaaS Service Disruption - June 26th, 2026
On June 26th, 2026 10:00AM PDT - 12:30PM PDT CVaaS service region cv-prod-us-central1-b experienced significant ingestion delay into NetDL that caused the following incorrect states to be observed by users:
During this time period, executing device provisioning actions may have been delayed or failed. There was no data loss. However, data ingested from actively streaming devices into CVaaS may have only become accessible after some delay. This may have impacted users accessing the UI, or other services consuming that data from CloudVision via API integrations.
Between June 26, 2026 12:30PM PDT - June 27th, 12:30PM PDT, some users may have experienced the above behaviors intermittently impact their devices, but not block device provisioning activities.
By June 27th, 2026, all mitigations were in place, and we transitioned to mitigating intermittent instances of elevated tail latency along the data ingestion path into NetDL. Between June 27th, 2026 and July 6th, 2026 a small subset of users of the cv-prod-us-central1-b service region may have experienced rare, but intermittent instances where CloudVision reported Event “High Streaming Latency”, potentially due to brief delays in NetDL ingestion. Users were not blocked from performing device provisioning activities.
Root Cause
Network performance degradation and instabilities in our underlying cloud provider were the ultimate root causes of delays ingesting data into the part of the system responsible for storing, processing, and updating device streaming status and streaming latency. Our mitigation activities between June 26th, 2026 and June 27th, 2026 stabilized the vast majority of the impact, but the changing underlying conditions prolonged the total duration of incident. This incident was not caused by software change(s), monitoring gaps, or software bugs in CloudVision as-a-Service.
Our Response
Multiple teams at Arista engaged around the clock from June 26th, 2026 until July 6th, 2026 to ensure we fully resolved this incident, as well as reviewed all details in our internal reviews. We continue to actively monitor this service region, as well as all of our CloudVision as-a-Service service regions.
Arista takes full responsibility for the performance and availability of the service regardless of the underlying cloud platform conditions. We are committed to making changes in both the short and long term that will provide greater resilience to underlying provider instabilities and changing real-time conditions.
Thank you for using CloudVision as-a-Service.