Global prune 2 #612
Global prune 2 #612
Conversation
dturn
force-pushed
the
global-deploy-2
branch
from
62c4cdc
to
3cc40d8
Compare
For sanity checking purposes, can you please post what this list actually is in the oldest and newest k8s versions we currently support? |
|
We intentionally black list Node and Namespace. I can't really see a reason where pruning a Node is a good idea. Pruning a Namespace is more reasonable but really dangerous. 1.11.10"/v1/PersistentVolume", 1.16.1"/v1/PersistentVolume", |
dturn
force-pushed
the
global-deploy-2
branch
2 times, most recently
from
7b616f1
to
fa17169
Compare
dturn
force-pushed
the
global-prune-2
branch
2 times, most recently
from
d4244b8
to
7c85a9f
Compare
| @@ -116,4 +126,54 @@ def test_global_deploy_task_failure | |||
| "Template validation failed", | |||
| ]) | |||
| end | |||
|
|
|||
| def test_global_deploy_prune_black_box_success | |||
| assert_deploy_success(deploy_global_fixtures('globals', clean_up: false, selector: 'test=prune1')) | |||
There was a problem hiding this comment.
same question, should we be adding this selector to the generated one so we don't have to hope future contributors won't reuse this manual selector?
| "Successful resources", | ||
| "StorageClass/testing-storage-class", | ||
| ]) | ||
| assert_deploy_success(deploy_global_fixtures('globals', selector: 'test=prune2')) |
There was a problem hiding this comment.
we should refute that the logs include "[pP]rune[d]"
There was a problem hiding this comment.
We use the test name to modify the fixture name, so if we refute 'prune' we can't use 'prune' in the test name. Instead I'm refuting 'pruned'.
dturn
force-pushed
the
global-prune-2
branch
from
5310285
to
afcff78
Compare
| @@ -103,6 +103,14 @@ def build_storage_v1_kubeclient(context) | |||
| ) | |||
| end | |||
|
|
|||
| def build_scheduling_v1beta1_kubeclient(context) | |||
There was a problem hiding this comment.
Is this actually necessary? I tried removing this and running tests but they keep passing; I tested with the tests modified/added in this PR:
$ alias rt="bundle exec ruby -I test"
$ rt test/integration-serial/serial_deploy_test.rb --name test_global_deploy_prune_black_box_success
$ rt test/integration/global_deploy_test.rb
$ rt test/unit/cluster_resource_discovery_test.rbAm I missing anything?
There was a problem hiding this comment.
build_scheduling_v1beta1_kubeclient are mostly used in the tests to allow for manual interaction with resources, usually to delete them in an ensure block. They aren't being used but seemed less than ideal to add a new kind to the fixtures but not add its helper methods.
What are you trying to accomplish with this PR?
Add pruning by default to global deploy
How is this accomplished?
Fetch a list of global resources that are pruneable (everything that supports delete) exclude node and namespace since I can't image we'd actually want to ever prune these and then pass them as the list of resources to prune.
What could go wrong?
Worst case is that we are including something in the prune list we shouldn't