Bug 1293744 - Extend the number of retries for taskgraph scheduling for release promotion. As a stopgap to help avoid failures. r=rail
authorJustin Wood <Callek@gmail.com>
Thu, 11 Aug 2016 10:30:20 -0400
changeset 6952 8686d725cd8c93eddbfd7ec1fb61b1fa926ecec0
parent 6950 f6c55a97b7054cbbaf838c3c403b1d98f1c34bd2
child 6953 31baa29d01d6f19c2b44b9cde6e28f7fe1719362
push id5186
push userCallek@gmail.com
push dateThu, 11 Aug 2016 14:56:28 +0000
reviewersrail
bugs1293744
Bug 1293744 - Extend the number of retries for taskgraph scheduling for release promotion. As a stopgap to help avoid failures. r=rail MozReview-Commit-ID: LQLCGsjDCyS
buildfarm/release/release-runner.py
--- a/buildfarm/release/release-runner.py
+++ b/buildfarm/release/release-runner.py
@@ -253,28 +253,35 @@ def main(options):
     smtp_server = get_config(config, 'release-runner', 'smtp_server',
                              'localhost')
     tc_config = {
         "credentials": {
             "clientId": get_config(config, "taskcluster", "client_id", None),
             "accessToken": get_config(config, "taskcluster", "access_token", None),
         }
     }
+    # Extend tc_config for retries, see Bug 1293744
+    # https://github.com/taskcluster/taskcluster-client.py/blob/0.0.24/taskcluster/client.py#L30
+    # This is a stopgap until Bug 1259627 is fixed.
+    retrying_tc_config = tc_config.copy()
+    retrying_tc_config = retrying_tc_config.update({
+        "maxRetries": 12
+    })
     configs_workdir = 'buildbot-configs'
     balrog_username = get_config(config, "balrog", "username", None)
     balrog_password = get_config(config, "balrog", "password", None)
     extra_balrog_submitter_params = get_config(config, "balrog", "extra_balrog_submitter_params", None)
     beetmover_aws_access_key_id = get_config(config, "beetmover", "aws_access_key_id", None)
     beetmover_aws_secret_access_key = get_config(config, "beetmover", "aws_secret_access_key", None)
     gpg_key_path = get_config(config, "signing", "gpg_key_path", None)
 
     # TODO: replace release sanity with direct checks of en-US and l10n revisions (and other things if needed)
 
     rr = ReleaseRunner(api_root=api_root, username=username, password=password)
-    scheduler = Scheduler(tc_config)
+    scheduler = Scheduler(retrying_tc_config)
     index = Index(tc_config)
     queue = Queue(tc_config)
 
     # Main loop waits for new releases, processes them and exits.
     while True:
         try:
             log.debug('Fetching release requests')
             rr.get_release_requests()