The Spanner team has recently been through an effort to make the timeout and retry settings consistent across all client library languages. The result of which is now in spanner_client_config.json.

We need to set different timeout_millis for each RPC method. But we also set specific retry settings for certain groups of methods.

We found that in gax-nodejs, if retry settings are set, then timeout_millis is ignored.

• https://github.com/googleapis/gax-nodejs/blob/aa5487b97f0346d0e4c3e67f0c1aedea2f128de5/src/gax.ts#L158-L159
• https://github.com/googleapis/gax-nodejs/blob/aa5487b97f0346d0e4c3e67f0c1aedea2f128de5/src/createApiCall.ts#L86-L99

This essentially means that all of our methods have a timeout_millis of 60 secs now instead of the method-specific ones we intended to set.

Would this be something that can be fixed in Node.js to respected the per-method timeout_millis?

CC @hengfengli