Upgrading retry-request to 4.2.2 makes some requests take a lot longer

I involuntarily upgraded to retry-request@4.2.2 (through @google-cloud/common@3.7.0 and others) and noticed that some of my tests were resulting in timeouts. It turns out that some operations take a lot longer (but don’t hang, it’s just that my test had a maximum running time of 5 seconds).

For example, trying to read from an index that doesn’t exist using a snapshot will take several seconds before failing, which isn’t the case with retry-request@4.2.0.

await snapshot.read(table.name, { index: "nope" });

Environment details

• OS: (At least) macOS
• Node.js version: v14.17.0
• npm version: 7.19.1
• @google-cloud/spanner version: 5.12.0

Steps to reproduce

1. Clone https://github.com/flovouin/spanner-retry-request-issue
2. Install dependencies
3. Ensure Docker is running (so that the Spanner emulator can be started)
4. Run ./run-test.sh 4.2.0 && ./run-test.sh 4.2.2

You should get something like this:

Using retry-request 4.2.0
5 NOT_FOUND: No index 'nope' was found for table 'test'.
Took 21 ms
Using retry-request 4.2.2
5 NOT_FOUND: No index 'nope' was found for table 'test'.
Took 6745 ms

Let me know if you need additional details. 🙂

Cheers,

[ansh0l]

@flovouin : This seems to have fallen through the cracks. Apologies for that. @asthamohta will take a look.

[skuruppu]

Thanks for the report @flovouin.

Since @stephenplusplus maintains retry-request, I am assigning this to him. If you can please take a look, and see if the issue is specific to Spanner or if it’s a generall issue, that would be great.