gRPC error spikes
Rails occasionally returns 502 and 503 HTTP status codes for various APIs kas
is calling. That results in kas
returning gRPC Unavailable
. That affects our gRPC error ratio SLI and is not great.
Filtering by json.err = "HTTP Status code:"
over the last 48 hours:
sum by (grpc_code,grpc_method,grpc_service) (
increase(grpc_server_handled_total{app="kas",env="gprd",grpc_code!="OK",grpc_code!="Canceled",grpc_code!="NotFound",grpc_code!="FailedPrecondition",grpc_code!="Unauthenticated",grpc_code!="ResourceExhausted"}[5m])
)