Use sort_keys with json almost everywhere we write to ZK

For almost any data we write to ZK (except for long-standing nodepool
classes), add the sort_keys=True so that we can more easily determine
whether an update is required.

This is in service of zkobject, and is not strictly necessary because
the json module follows dict insertion order, and our serialize methods
are obviously internally consistent (at least, if they're going to produce
the same data, which is all we care about).  But that hasn't always been
true and might not be true in the future, so this is good future-proofing.

Based on a similar thought, the argument is also added to several places
which do not use zkobject but which do write to ZK, in case we perform
a similar check in the future.  This seems like a good habit to use
throughout the code base.

Change-Id: Idca67942c057ab0e6b629b50b9b3367ccc0e4ad7

1 files changed, 2 insertions, 1 deletions

diff --git a/zuul/executor/server.py b/zuul/executor/server.py
index fc7760472..204bbad03 100644
--- a/zuul/executor/server.py
+++ b/zuul/executor/server.py
@@ -1226,7 +1226,8 @@ class AnsibleJob(object):
                         result=None,
                         error_detail=f'Failed to update project '
                                      f'{task.project_name}')
-                    self.job.sendWorkComplete(json.dumps(result))
+                    self.job.sendWorkComplete(
+                        json.dumps(result, sort_keys=True))
                     return
 
                 raise ExecutorError(