stacklok / toolhive / 20961566455

58%

Build Type

push

github

Committed by web-flow

Commit Message

Fix: Migrate telemetry config `samplingRate` from float64 to string (#3268)

Users who created workloads with telemetry enabled in older versions of ToolHive encounter noisy warnings and errors when upgrading:

WARN  failed to load state for fetch: json: cannot unmarshal number into Go struct field Config.telemetry_config.samplingRate of type string
This occurs because the SamplingRate field in telemetry.Config was changed from float64 to string, but existing run config files still contain the old numeric format (e.g., "samplingRate": 0.1 instead of "samplingRate": "0.1").

Fixes #3267

Solution
Added a migration that runs at startup to convert old run config files from the float format to the string format.

The migration:

Runs once per installation (tracked by TelemetryConfigMigration flag in config)
Lists all run config files and checks if telemetry_config.samplingRate is a number
Converts numeric values to strings and writes them back
Logs which workloads were migrated
Is idempotent (safely skips configs that already have string values or no telemetry config)
Changes
cmd/thv/main.go: Added call to telemetry config migration at startup
pkg/config/config.go: Added TelemetryConfigMigration flag to track migration completion
pkg/migration/telemetry_config.go: New migration implementation
Testing
Before migration:

$ ./bin/thv-old run --name fetch --otel-endpoint http://localhost:4318 --otel-sampling-rate 0.1 fetch
# Creates config with: "samplingRate": 0.1

$ go run ./cmd/thv restart fetch
WARN  failed to load state for fetch: json: cannot unmarshal number into Go struct field Config.telemetry_config.samplingRate of type string
Error: failed to restart workload fetch: ...
After migration:

$ go run ./cmd/thv restart fetch
INFO  Successfully migrated telemetry config for 1 workload(s)
INFO  Loaded configuration from state for fetch
INFO  Starting tooling server fetch...
Workload fetch restarted successfully
# Config now has: "samplingRate": "0.1"

Run Details

45 of 141 new or added lines in 3 files covered. (31.91%)

32 existing lines in 3 files now uncovered.

34424 of 59368 relevant lines covered (57.98%)

81.8 hits per line