update the info on the Eviction API (node drain)

Author: mimowo

keps/sig-apps/3329-retriable-and-non-retriable-failures/README.md

@@ -92,9 +92,9 @@ tags, and then generate with `hack/update-toc.sh`.
     - [Current state review](#current-state-review)
       - [Preemption](#preemption)
       - [Taint-based eviction](#taint-based-eviction)
+      - [Node drain](#node-drain)
       - [Node-pressure eviction](#node-pressure-eviction)
       - [OOM kill](#oom-kill)
-      - [Node drain](#node-drain)
       - [Disconnected node](#disconnected-node)
       - [Direct container kill](#direct-container-kill)
     - [Termination initiated by Kubelet](#termination-initiated-by-kubelet)
@@ -486,6 +486,22 @@ be retried from those which should not be retried.
   - `reason=Error`
 - Retriable: Yes
 
+##### Node drain
+
+- Reproduction: We run a job with a long-running pod, then drain the node
+  with the `kubectl drain` command
+- Comments: performed by Eviction API, controlled by kube-apiserver in `registry/core/pod/storage/eviction.go`
+- Pod status:
+  - status: Terminating
+  - `phase=Failed`
+  - `reason=`
+  - `message=`
+- Container status:
+  - `state=Ternminated`
+  - `exitCode=137`
+  - `reason=Error`
+- Retriable: Yes
+
 ##### Node-pressure eviction
 
 Memory-pressure eviction:
@@ -540,22 +556,6 @@ Disk-pressure eviction:
   `resources.requests.memory=resources.limits.memory` it strongly suggests
   a software bug.
 
-##### Node drain
-
-- Reproduction: We run a job with a long-running pod, then drain the node
-  with the `kubectl drain` command
-- Comments: performed by Eviction API
-- Pod status:
-  - status: Terminating
-  - `phase=Failed`
-  - `reason=`
-  - `message=`
-- Container status:
-  - `state=Ternminated`
-  - `exitCode=137`
-  - `reason=Error`
-- Retriable: Yes
-
 ##### Disconnected node
 
 - Reproduction: We run a job with a long-running pod, then disconnect the node