Eventarc trigger vs HTTP trigger performance

[Karthik Katta]

Hi,

We are seeing a strange behavior.  If a cloud function configured as "Eventarc trigger" is running too slow vs function configured as "HTTP trigger". 

The source code is simple and same between both these functions. The code just downloads a file from amazon s3.

While the "HTTP trigger" downloads the file in few seconds, the "Eventarc trigger" function is taking minutes to download the file. I confirmed that resources allocated for both these functions are same (32 GB RAM and 8 cores) which is more than enough for this simple code.

Any pointers on this would be really helpful. Thank you all.

[Jeff Huleatt]

Hi Karthik,

That's an interesting problem! As an early guess, if your HTTP trigger is hit more consistently, it could be running off of a warm instance, while your Eventarc trigger could be scaling to zero and then cold starting every time it is called. Any chance that could be happening? Otherwise, are there any differences between the code inside the functions, such as where variables are initialized?

If those two don't help, are you able to share the code from these functions, preferably pared down to just enough code to demonstrate the problem? There are a number of things that could be going on, and some code to help reproduce the problem would help narrow things down.

Best,

Jeff