10

I am using nodejs on appengine. We had a perfectly stable app scaffold on tuesday, and by friday it was completely broken after attempting to deploy. We made some minor changes to frontend code, but nothing that I think would prevent the instance from spinning up.

Here are steps to reproduce.

  1. gcloud --project "{appname}" preview app deploy

  2. logs show npm install, container build, etc.

  3. Hangs on Updating service [default]... for 5 minutes

  4. Fails with error.

ERROR: (gcloud.preview.app.deploy) Error Response: [13] Timed out when starting VMs. It's possible that the application code is unhealthy. (0/1 ready, 1 still deploying).

I have tried reverting the repository to when we had stable deployments, and it didn't help. This makes me think something on GCP is broken.

I have tried deleting all current versions and then deploying, but to no avail.

Improve this question

3 Answers 3

Reset to default
4

When you get this error, you can take a look at crash.log in the Cloud Console Logs Viewer (Logging -> Logs) for your specific service / version which will usually tell you exactly what happened. In my case when I reproduced the same error, crash.log shows me the output of NPM which tells me I have a 'SyntaxError: Unexpected identifier'.

YMMV of course, but this can tell you if the issue is related to your application code or if there is something more sinister going on.

Improve this answer
3
  • Thanks, I'll try that next time around. I had to abandon the appengine approach due to short timeline. Is there some way to build and run the container locally?
    – posit labs
    May 29, 2016 at 3:05
  • It's not necessary to build the container for local testing of nodejs apps, you can just run them locally as you would any nodejs app with 'npm install; npm start'.
    – Adam
    May 29, 2016 at 18:03
  • I started from scratch and things seemed to work the second time around. No change in the app code, but I downgraded to node 4.4.4 in package.json's engines field.
    – posit labs
    Jun 11, 2016 at 1:38
2

The answer for me was different... The version started to get created but never finished the creation process. This is scenario is visible if you go into the logs for your AppEngine service (check all options to see ALL logs) and you may notice "Container called exit(1)."

In my case I found that running "gcloud init" to reset my credentials helped. Hope this helps someone else.

Scratch that... I found that the Cloud Build API was just WAY behind / slow. This has been happening for the last couple of hours. I just noticed 4 versions pop into my dashboard all at once from prior hours in the evening. Apparently there is no way to cancel prior deployments? So ironically once Google's build service gets behind everyone probably starts spamming / retrying their deployments and the issue gets worse... so bad.

Improve this answer
1

Check to make sure your package.json has this section

"cloud-repo-tools": {
"test": {
  "app": {
    "msg": "Hello, world!"
  }
},
"requiresKeyFile": true,
"requiresProjectId": true
}

with the "msg" section containing some string that the health check can look for. Haven't been able to find documentation for this, so if anyone else does I'd love to see it.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .