Issues with Build Timeout

[mitchellmaciorski]

I'm trying to build an application in a python flex environment that is using some fairly large files (I'd say in total somewhere around 3GB). I was getting some timeout errors which i fixed through increasing the build timeout parameter, but at the final stage now I'm getting a new error:

Updating service [(myservice_name)] (this may take several minutes)...failed.                               

ERROR: (gcloud.app.deploy) Operation [apps/(myproject_name)/operations/0ee990a6-6ddb-4bc4-ba1e-80b309777866] timed out. This operation may still be underway.

I've tried increasing the app_start_timeout_sec parameter to it's maximum allowed value (3600), but it doesn't seem to have helped. Any help / suggestions would be greatly appreciated :)

[mitchellmaciorski]

Update:

I realized that only one of the two (groups of) files was quite so large, so I retried deploying with ~150MB of files involved. Unfortunately, it didn't help.

[Kenworth (Google Cloud Platform)]

For deployments, there is a limit of 32MB for each file, 10k uploaded files per version, and free 1GB total size of all files for all versions. Make sure you are not hitting these limits. 

Moreover, can you clarify ~3GB? Is this deployment size? If yes, this is pretty big imho.

[mitchellmaciorski]

Well, that clarifies things. I've been using app engine to deploy small machine learning models and I just switched over to some larger ones, so I guess that's why I've never run into this issue before. I'm not using tensorflow models - is there a google cloud service that would be better suited to handling files of this size?

[mitchellmaciorski]

Doesn't the 32MB per deployment limit only apply to the standard environment? I have another project deployed (python3 flex environent) that's currently in use with a 300MB .csv file included, so this doesn't actually make any sense.

[mitchellmaciorski]

Fixed the issue. You can deploy apps with files larger than 32MB in a flex environment. The problem was that I didn't provide a route in my app for a readiness check. What's weird is that I've never had an issue with that before. Maybe it was a legacy / updated health checks thing. I'm pretty sure all my other deployments have used legacy health checks.