在cron.yaml上设置retry_parameters

Question

I am setting up a cron job related to an App Engine app (standard environment, Python 3) and want it to retry after 2 minutes on failure. No matter what value I put in retry_parameters it seems to be retrying in 1 minute. Looking at the docs, I don't see a mention of a max wait time, so I wonder if I've just messed up the cron.yaml somehow.

Everything is working perfectly, except it is retrying sooner than I would like. Here's the yams — is this a limit or have messed something up.

cron.yaml :

cron:
- description: "daily call"
  url: /twilio/start_call/
  timezone: "America/Anchorage"
  schedule: every day 17:00
  retry_parameters:
    min_backoff_seconds: 120.0
    max_backoff_seconds: 360.0
    max_doublings: 3

Edit:

Here's a sample of the Stackdriver logs generated with the cron.yaml . The function returned 503 each time until the last and cron was firing the jobs at 1-minute intervals:

2018-11-26 17:00:00.764 AKST GET 503 178 B 970 ms AppEngine-Google (+http://code.google.com/appengine) /twilio/start_call/
2018-11-26 17:01:01.939 AKST GET 503 178 B 704 ms AppEngine-Google;(+http://code.google.com/appengine) /twilio/start_call/
2018-11-26 17:02:02.747 AKST GET 503 178 B 850 ms AppEngine-Google;(+http://code.google.com/appengine) /twilio/start_call/
2018-11-26 17:03:03.702 AKST GET 503 178 B 666 ms AppEngine-Google;(+http://code.google.com/appengine) /twilio/start_call/
2018-11-26 17:04:04.477 AKST GET 200 189 B 65 ms AppEngine-Google; (+http://code.google.com/appengine) /twilio/start_call/

Here's what a particular log entry looks like expanded:

 { httpRequest: { status: 503 } insertId: "5bfca521000b366ac405955e" labels: { clone_id: "00c61b117cda9e441fb74ae3fd3225f528216a2a7e2fd701de97c95ac79fdc86ce3014f397" } logName: "projects/caller-app/logs/appengine.googleapis.com%2Frequest_log" operation: { first: true id: "5bfca52000ff0baa94727d8d2500016d7e7368656c7465722d63616c6c65720001323031383131323674313033313038000100" last: true producer: "appengine.googleapis.com/request_id" } protoPayload: { @type: "type.googleapis.com/google.appengine.logging.v1.RequestLog" appEngineRelease: "1.9.65" appId: "m~caller-app" cost: 1.9892999999999998e-8 endTime: "2018-11-27T02:00:01.734682Z" finished: true first: true host: "caller-app.appspot.com" httpVersion: "HTTP/1.1" instanceId: "00c61b117cda9e441fb74ae3fd3225f528216a2a7e2fd701de97c95ac79fdc86ce3014f397" instanceIndex: -1 ip: "0.1.0.1" latency: "0.970118s" megaCycles: "123" method: "GET" requestId: "5bfca52000ff0baa94727d8d2500016d7e7368656c7465722d63616c6c65720001323031383131323674313033313038000100" resource: "/twilio/start_call/" responseSize: "178" startTime: "2018-11-27T02:00:00.764564Z" status: 503 taskName: "22a1b20373e6b0b93d21726ad7218cff" taskQueueName: "__cron" traceId: "54903137240fb57e546907087fb94ca5" traceSampled: true urlMapEntry: "auto" userAgent: "AppEngine-Google; (+http://code.google.com/appengine)" versionId: "20181126t103108" } receiveTimestamp: "2018-11-27T02:00:01.759151584Z" resource: { labels: { module_id: "default" project_id: "caller-app" version_id: "20181126t103108" zone: "us-west2-3" } type: "gae_app" } timestamp: "2018-11-27T02:00:00.764564Z" trace: "projects/caller-app/traces/54903137240fb57e546907087fb94ca5" traceSampled: true } 

Answer 1

So this was a simple case of mis-reading the documentation. But in case anyone misreads it as well…

Responding with a 503 is a special case. The documentation explains:

By default, failed jobs are not retried unless a 503 status code is returned, in which case it is retried every minute until it succeeds or returns a 200-299 status code.

It doesn't explicitly say it, but this means that when the server responds with a 503 status the retry_parameters in cron.yaml are ignored and it still retries every minute.