docs: iteration on aws k8s upgrade docs

docs/Makefile

@@ -14,8 +14,9 @@ help:
 
 .PHONY: help Makefile
 
+# keep --ignore entries in sync with content in noxfile.py
 live:
-	sphinx-autobuild --ignore "_build/*" --ignore "tmp/*" --ignore "_static/hub-*.json" -b dirhtml -n . _build/dirhtml
+	sphinx-autobuild --ignore "*/_build/*" --ignore "*/tmp/*" --ignore "*/*.json" --ignore "*/*.csv" -b dirhtml -n . _build/dirhtml
 
 # Catch-all target: route all unknown targets to Sphinx using the new
 # "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).

docs/howto/upgrade-cluster/index.md

@@ -50,6 +50,7 @@ some in shared clusters have been announced ahead of time.
 :maxdepth: 1
 :caption: Upgrading Kubernetes clusters
 upgrade-disruptions.md
+node-upgrade-strategies.md
 k8s-version-skew.md
 aws.md
 ```

docs/howto/upgrade-cluster/node-upgrade-strategies.md

@@ -0,0 +1,25 @@
+(upgrade-cluster:node-upgrade-strategies)=
+
+# About strategies to upgrade nodes
+
+## About rolling upgrades
+
+To upgrade node group's nodes, we typically do *rolling upgrades*. In a rolling
+upgrade, something new is added to replace something old before its removed.
+
+When doing a rolling upgrade of node groups, we can do a rolling upgrade _fast
+and forcefully_ or _slow and patiently_ - either pods running on a node group's
+nodes get forcefully stopped, or they get to stop on their own.
+
+*Managed* node groups can do fast and forceful rolling upgrades, while
+*unmanaged* node groups need to be re-created to get upgraded k8s software
+(`kubelet` etc).
+
+Core nodes' workloads can be suitable to forcefully relocate, while user nodes'
+workloads should be given time to stop on their own.
+
+## About re-creation upgrades
+
+With unmanaged node groups like on EKS, if disruption isn't a concern or if
+there isn't anything running to disrupt, node groups can be deleted and
+re-created to save time.

docs/hub-deployment-guide/new-cluster/new-cluster.md

@@ -35,8 +35,13 @@ This guide will assume you have already followed the guidance in [](/topic/infra
 
    Mac users with homebrew can run `brew install eksctl`.
 
-   Verify install and version with `eksctl version`. You typically need a very
-   recent version of this CLI.
+   Verify install and version with `eksctl version`. You should have *the latest
+   version* of this CLI.
+
+   ```{important}
+   Without the latest version, you may install an outdated versions of `aws-node`
+   because [its hardcoded](https://github.com/eksctl-io/eksctl/pull/7756).
+   ```
 
 4. Install [`jsonnet`](https://github.com/google/jsonnet)
 

eksctl/2i2c-aws-us.jsonnet

@@ -86,7 +86,7 @@ local daskNodes = [
     nodeGroups: [
         ng + {
             namePrefix: 'core',
-            nameSuffix: 'b',
+            nameSuffix: 'a',
             nameIncludeInstanceType: false,
             availabilityZones: [nodeAz],
             ssh: {

noxfile.py

@@ -26,13 +26,16 @@ def docs(session):
         session.posargs.pop(session.posargs.index("live"))
 
         # Add folders to ignore
+        # keep this in sync with Makefile
         AUTOBUILD_IGNORE_DIRS = [
             "_build",
             "tmp",
         ]
         # Add files to ignore
+        # keep this in sync with Makefile
         AUTOBUILD_IGNORE_FILES = [
-            "_static/*.json",
+            "*.json",
+            "*.csv",
         ]
 
         cmd = ["sphinx-autobuild"]

terraform/gcp/projects/qcl.tfvars

@@ -26,7 +26,7 @@ user_buckets = {
 }
 
 notebook_nodes = {
-  # FIXME: tainted, to be deleted when empty, replaced by k8s upgraded variant
+  # FIXME: tainted, to be deleted when empty, replaced by equivalent during k8s upgrade
   "n2-highmem-4" : {
     min : 0,
     max : 100,