Bug #17561
Updated by Tom Clegg almost 3 years ago
Currently, a-d-c and its drivers detect when the cloud provider returns a "too many requests" response to a "list instances" or "create instance" API call, and waits for some time to elapse before trying the same type of call. We should also detect & handle those responses for "update instance tags" and "destroy instance" calls. a-d-c's behavior will seem a bit strange if the sysadmin doesn't realize these calls are being rate-limited. It might be worth adding a metric like @arvados_dispatchcloud_ratelimit{api="ListInstances"}=1@ when limited, 0 when not.