Double-counting requests if using vhost functionality?

[aseemk]

Hi guys,

We're running a Node.js app on Heroku and have node-newrelic capturing metrics. We've noticed that New Relic's count of requests per minute is consistently ~2x what we're seeing from Heroku's router logs (e.g. via log2viz, but also in Librato via l2met).

(I wish I could check Nodetime's count too. Unfortunately at the moment their site seems to be broken. I'll update this whenever it works again and I can check.)

It occurred to me that what could be happening is that requests might be getting double-counted, since we use Connect's/Express's vhost functionality, so all of our requests technically flow through two Express servers.

Simplified code: https://gist.github.com/aseemk/5613063

Is that a possibility? If so, it'd be great to fix.

We're still on Express 2 at the moment (and Node 0.6), but are actively looking to upgrade to Express 3 (and Node 0.10) very soon.

Thanks much!

[othiym23]

You're right that it shouldn't be double-counting like that. We'll take a look and let you know what we find. Thanks for the report and the code sample!

[othiym23]

Fixed via 209247e.

[aseemk]

Deployed 0.11.2 to production and we're still seeing this FYI! Not urgent for us by any means, so don't feel the need to worry about this, but just wanted to let you know.

[othiym23]

Weird! The changes I made to the tracer should have put a stake through this. Can you put together a reduced test case for this so we can be sure we've killed the bug for good? It doesn't have to be pretty, it just has to reliably demonstrate the problem. Thanks, and sorry for closing this when it wasn't fixed. :/

[othiym23]

Give 0.11.4 a shot and see if it makes a difference. It may not, but the change I made to the Connect instrumentation may affect this too.

[seth-admittedly]

I'm having a related problem with mounted apps. I've got both the relative and absolute routes ignored in my config file, but the requests are still showing in my reports and the logger is spitting out {"name":"newrelic","hostname":"local.admitted.ly","pid":20481,"component":"express","level":50,"msg":"No Express route to use for naming.","time":"2013-12-12T06:17:37.844Z","v":0} for each request.

[othiym23]

👍 This is something I want us to get to soon, as part of a larger rewrite of router introspection to be more general. Thanks for moving this over here, @seth-admittedly!

[txase]

Hi @seth-admittedly and @aseemk,

This issue is a bit stale, and it's not clear if there is still an issue in the latest version of the agent. Please give it a try.

I am going to close this issue under the assumption that we have fixed the issue, but if not please feel free to reopen the issue. Thanks!

[aseemk]

Sorry for the delay here guys, and thanks for the work!

We are going to be upgrading our node-newrelic soon and will report back!