Prevent API groups from growing stale #264
Conversation
openshift-ci-robot
added
the
size/S
|
What is the difference between the existing refresh mechanism and the one resulting from this change? It's not obvious to me how the refresh is triggered. |
openshift/dynamic/client.py
Outdated
| if request_resources: | ||
| resources = self.get_resources_for_api_version(DISCOVERY_PREFIX, group['name'], version, preferred) | ||
| new_group[version] = ResourceGroup(preferred, resources=resources) | ||
| groups[DISCOVERY_PREFIX][group['name']] = new_group | ||
| self._cache['resources'] = groups | ||
| self._cache['resources'].update(groups) |
There was a problem hiding this comment.
Wouldn't doing the dictionary update every iteration of the loop be slower than just setting self._cache['resources'] at the end?
There was a problem hiding this comment.
woops, typo, that was supposed to be one level out
|
This makes the api_groups refresh when accessed, so that it's always the most up to date list of resources available in the cluster. Before it would only refresh when you attempted to access a resource that did not exist in the cache. Since updating api_groups is just a single API call (with the lazy discoverer at least), I thought this behavior would be more useful than requiring you to manually refresh the cache or miss a resource to get the new groups to show up. We hit this in one of our operators, where it created a CRD and later verified that the API group existed- because api_groups didn't refresh, the check failed. |
|
I think I tested this correctly: Then I create my CRD |
|
And I can show it failing on |
|
@willthames did that clear up your reservations? Any opposition to merge? |
* Make api_groups a property * Prevent api_groups from growing stale by re-requesting resources every time * Add back v1.List * Typo
@willthames @asetty if you'd like to take a look