Handshake Inactivity Timeout - PROTOCOL_SEQUENCE_TIMEOUT in 2.4.1

[inspectlet]

I'm upgrading from 2.0.0-alpha9 to 2.4.1 and I'm seeing a "handshake inactivity timeout" fatal error with code "PROTOCOL_SEQUENCE_TIMEOUT", any idea why this could be happening?

Some background: our servers are on Heroku using AWS RDS.

I've seen that these errors happen at least when Heroku dynos are restarted (like after a deploy), and continue to happen for about 2 minutes until they die down. I'm also using the connection pooling and give each worker in the cluster a pool of up to 5 nodes.

This morning we saw that the MySQL DBs had completely blocked Heroku hosts, probably due to very frequent reconnection. I've since downgraded back to 2.0.0-alpha9 and I'm unable to reproduce the problem locally.

Any help figuring out what's going on is greatly appreciated :)

[dougwilson]

Hi!

So back in the old versions, there were no timeouts at all. Now that there are actually timeouts, you are just bumping into one. What you'll want to do is adjust your pool config option for acquireTimeout to something higher.

[purejs]

Seems the unit of acquireTimeout is millisecond rather than second. So the doc somehow is misleading by the description of its default value.

[dougwilson]

Ho so? The very first words are "The milliseconds before a timeout occurs during the connection acquisition" stating that it's milliseconds. The default value is easier to read as "10 seconds" instead of "10000 milliseconds", and it's also not a literal value, since it's not in a code block like the other defaults are.

[onury]

I agree with @purejs. Turns out, I've configured the value 10 bec. of the docs. What's hard to read with "10000 milliseconds"?

[dougwilson]

Ok, I have just updated the documentation to show the literal default value so you no longer need to read the docs to see the value takes milliseconds.