downward API and parallelism

Signed-off-by: Aldo Culquicondor <acondor@google.com>
kubernetes · Jan 21, 2021 · 14dc865 · 14dc865
1 parent d12b83e
commit 14dc865
Showing 1 changed file with 9 additions and 8 deletions.
diff --git a/keps/sig-apps/2214-indexed-job/README.md b/keps/sig-apps/2214-indexed-job/README.md
@@ -14,7 +14,7 @@
 - [Design Details](#design-details)
   - [JobSpec API](#jobspec-api)
   - [Job completion and restart policy](#job-completion-and-restart-policy)
-  - [Job scaling](#job-scaling)
+  - [Job parallelism](#job-parallelism)
   - [Test Plan](#test-plan)
   - [Graduation Criteria](#graduation-criteria)
     - [Alpha -&gt; Beta Graduation](#alpha---beta-graduation)
@@ -226,7 +226,8 @@ spec:
 ```
 
 The Job controller doesn't add the environment variable if there is a name
-conflict with an existing environment variables in the Job's Pod template.
+conflict with an existing environment variables in the Job's Pod template or if
+the user already specified another environment variable for the same annotation.
 
 ### Job completion and restart policy
 
@@ -272,14 +273,14 @@ Completed Indexes: [1-25,28,30-32,...]
 ``` 
 <<[/UNRESOLVED]>>
 
-### Job scaling
+### Job parallelism
 
-Jobs can be scaled up and down by varying `.spec.parallelism` (note that
-`.spec.completions` is an immutable field).
+A user can change the number of active Pods for a Job changing `.spec.parallelism`
+(note that `.spec.completions` is an immutable field).
 
-When starting or scaling up a Job, the Job controller creates Pods with lower
-completion index first, as long as there is no other completed or running Pod
-with the same index. This is to make the controller behavior more
+When starting a Job or increasing the parallelism, the Job controller creates
+Pods with lower completion index first, as long as there is no other completed
+or running Pod with the same index. This is to make the controller behavior more
 predictable. We do not offer guarantees on creation order based on completion
 index.