Werk #22297: Azure Cosmos DB: container services of idle containers no longer vanish
| Component | Core & setup | ||||
| Title | Azure Cosmos DB: container services of idle containers no longer vanish | ||||
| Date | Sep 16, 2026 | ||||
| Level | Trivial Change | ||||
| Class | Bug Fix | ||||
| Compatibility | Compatible - no manual interaction needed | ||||
| Checkmk versions & editions |
|
The services "Azure/Cosmos DB container name Requests" and "Azure/Cosmos DB container name Throughput" repeatedly vanished, turned UNKNOWN with "Item not found in monitoring data", and reappeared as new services a few minutes later. Containers with little or no traffic were affected most.
The agent derived databases and containers from Azure Monitor time series only. Azure returns no series for a container without requests in the queried window and, without an explicit limit, at most 10 series per request in no stable order. Both made containers disappear from the agent output between runs.
The agent now requests up to 50 time series per metric, ordered by value, and lists the databases and containers of SQL (NoSQL) and MongoDB accounts through the Azure resource provider. Every listed container gets a service. A container without requests reports "No requests" and stays OK. Databases without traffic now also appear as piggyback hosts.
The service principal needs read access to the database account and its databases and containers, which the Reader role includes.