- 2.8 based bug
- Only reproducible when
- performance settings have been changed
- navigating from performance page to list page
- Root effect
- Contstant trigger of two computed properties
- namespaceFilterRequired
- shell/mixins/resource-fetch-namespaced.js
- namespaceFilter --> __namespaceRequired --> value (all same colour, suggesting our code)
- tracing shows this also to start from shell/mixins/resource-fetch-namespaced.js
* Add tests for setting up vai env in rancher
* workflow hack
* testing
* add user tag to @vai tests
* more shippable trigger, in the hope it fixes the looong test run due to tests with no tests in...
* test - rely on filename sort order for priority
* test - rely on spec order for priority
* test - rely on spec order for priority - take 2
* add setup for vai...
* Apply tags to new tests
* fix global roles - cluster different if run directly
- there was a more complex solution here (hence abstract) but it ended up needing to go to a cluster page anyway, so might as well just nav to desired page
* fix users - cluster different if run directly
* Fix other revisions (probably more TODO...)
* Fixes for event and crd tables
* Tweaks
* Revert temp changes
---------
Co-authored-by: Richard <richard@linux.fritz.box>
* Changes for new design
- New visuals
- Pagination controls --> load more
- finished testing of label select with pagination off
# Conflicts:
# shell/edit/provisioning.cattle.io.cluster/__tests__/Basics.tests.ts
* Changes following review
* Update Node list to support server-side pagination
- Setup pagination headers for the node type
- Define a pattern for fetching custom list secondary resources
- Major improvements to the way pagination settings are defined and created
- Lots of docs improvements
- Handle calling fetch again once fetch is in progress (nuxt caches running request)
- Validate filter fields (not all are supported by the vai cache
- General pagination fixes
* Lint / test / fixes
* Improvements to configmap e2e test & Improve pagination disabled
* Beef up validation
* Fix missing name column in non-server-side paginated node list
* Fix PR automation actions
- fix syntax
- catch scenario where a pr has no fixed issue
> There's duplication between files, see https://github.com/rancher/dashboard/pull/10534
* CI bump
* Fixes post merge
* Wire in 2.9.0 settings for server-side pagination
- Everything is gated on `on-disk-steve-cache` feature flag
- There's a backend in progress item to resolve a `revision` issue, until then disable watching a resource given it
- Global Settings - Performance
- Added new setting to enable server side pagination
- this is incompatible with two other performance settings
* Integrate pagination with configmaps in cis clusterscanbenchmark edit form
Also
- improved labeled select pagination
- gate label select pagination functinality on steve cache being enabled
* TODOs and TEST
* Paginate Secret selection for logging providers
- Allow `None` option in Paginationed LabelSelect
- Optionally classify pagination response
* WIP
* fixes arfter merge
* Don't suggest container names, not practical
- previously all pods were fetched... and we scrapped all container names from them
- this is a scaling nightmare, user now must just enter the name/s to match
* Avoid findAll secrets in SimpleSecretSelector
* tidying up
* Move LabeledSelect/index.vue back to LabeledSelect.vue to not break extensions
* changes after self review... 1
* changes after self review... 2
* changes after self review... 3
* Link new paginated label select with pagination setting
* Work around failing kubewarden unit tests in check-plugins gate
* Fix backup.spec e2e test
* Create a convienence wrapper called ResourceLabelSelector that hides most of the complexity
* fix unit test
- Everything is gated on `on-disk-steve-cache` feature flag
- There's a backend in progress item to resolve a `revision` issue, until then disable watching a resource given it
- Global Settings - Performance
- Added new setting to enable server side pagination
- this is incompatible with two other performance settings
> This uses a new endpoint that has yet to merge. See https://github.com/rancher/rancher/issues/40140
WIP
- Contains console.warns (via custom logger, can be disabled)
- Waiting for final BE endpoint changes to merge
- Contains TODOs to resolve on final enpoint changes delivered
Pertinent Points
- Incompatible with incremental loading / manual refresh
- Harder to get counts (need to sum up from different namespaces)
- Requires use of new steve pagination
- Enforced NS threshold has been removed
- The threshold only applies to the primary resource. This has issues when loading a low count primary (daemon sets) which depends on a very high count secondary (pods)
- Fixing this would involve knowing all secondary resources a list uses, which isn't currently possible (each resource is requested individually, need to know them all first)
- There is no way to subscribe to multiple namespaces (one or all)
- We mock this in subscribe by only persisting changes to resources from within target namespaces
- Everything should work with Advanced Worker enabled
* Moves sockets into the advanced worker
* worker can die peacefully now, making switching between cluster work.
* Make waitFor generic, wire in to waitForTestFn
* General Changes
- Fixes for switching cluster
- includes using common getPerformanceSetting
- avoid new code to unsub before socket disconnect
- handle `watch` `stop` requests
- lots of TODO's (questions, work, checks, test, etc)
- use common
* Switch socket fixes
- isAdvancedWorker should only be true for cluster store
- advancedWorker to be wired in
* Fix socket id for cluster workers
- sockets use an incremented local var for id
- when we nuke the socket file within the worker this resets, so they all ahve id of 1
- work around this by applying the unix time
* Fix handling of new partical counts response
- seen in dex cluster explorer dashboard
- count cards would be removed when partial counts response received
* Make resourceWatcher the sole location for watch state
- getters canWatch, watchStarted now are worked around (they look at state in the UI thread)
- we now don't call resource.stop or restart.start in subscription
- tidied up `forgetType`
- moved clearFromQueue from steve mutations into subscription mutations (better location)
- added and removed some TODOs
- fixed watch (stop handler should be higher up, include force watch handling)
* pushes the csrf value into worker and adds it to fetch request headers.
* refactors batchChanges to address ref concerns and be more performant
* Maintain schema reference whilst updating
- This change mutates input in a function, which is bad...
- but ensures the reference isn't broken, which is needed to maintain similar functionality as before
* Fix waitForTestFn
- Seen when creating or viewing clusters
* On unwatch ensure any pending watch requests are removed from the queue
- the probably would have been a problem if the worker wasn't nuked
- however as the codes there lets make it safe
Also added `trace` feature in advanced worker, will probably bring out to other places as well
* Fix navigation from cluster manager world to any cluster
- Ensure that we handle the case where the advanced worker was created but the resource watcher wasn't
- ... but fix case where this was happening (aka ensure that a blank cluster context is ignored)
* Tidy some TODOs
* Add perf settings page
- This will help test normal flow (when advanced worker is disabled)
- Note - setting is now in a bag. This may help us better support further settings (enable client side pagination, etc)
```
advancedWorker: { enabled: false },
```
* FIX - Nav from cluster dashboard --> specific event --> cluster dashboard and events not re-subbed
- Ensure we block default handling of resource.start (keep state in resource watcher)
* Tidying up some TODOs
* Adds in a cache and uses it to validate SCHEMA messages before batching.
* Forgot to actually save CSRF to the resourceWatcher when instantiated.
* an empty resource in a batchChange to signal remove
* Move addSchemaIndexFields to and created removeSchemaIndexFields in new file
- this avoids bringing class files into the worker
* Fix disconnect/reconnect
- Remove `syncWatch` (do the watch/unwatch straight away)
- Test/Fix re-sub on reconnect
- Test/Fix growls on disconnect
* Tidying up some TODO's
- including clean of workerQueue on resource.stop (this is SUPER defensive)
* batchChanges will now handle aliases
* Fix pods list - WIP
- ensure podsByNamespace is updated on batchChange
TODO
- the final update to the pod is ignored
- removing a namespace cleans the cache correctly
- disabling advanced worker still works
* Fix pods list - fixes
- ensure podsByNamespace is updated on batchChange
Tested / Fixed
- the final update to the pod is ignored
- removing a namespace cleans the cache correctly
- disabling advanced worker still works
* Tidying TODOs
* Remove default same-origin header
- https://developer.mozilla.org/en-US/docs/Web/API/Request/credentials
* Fixed TODO description
* Refactor subscribe, make it clear which vuex feature relates to what
* Lots of Fixes
- batchChanges fixes
- fix index is 0 issues (!/!!index)
- only `set` if we have to
- ensure we set the correct index after pushing to list
- ensure map is updated after reducing list size with limit
- podsByNamespace fixes
- ensure when ew replace... we don't use the same referenced object
- general service resource fixes
- ensure service's pods list stays up to date with store
* Multiple improvements/fixes
- resourceCache - store the hash instead of the whole object. This means longer load time be reduces memory footprint
- resourceWatcher
- don't re-sub on socket reconnect if watcher is in error
- don't sub if watcher is in error
- don't unwatch for 'failed to find schema' and 'too old' errors
- this clears the error, we won't to keep it to ensure we don't watch
- Remove #5997 comments, follow on work #7917
* toggle debug, remap alias types, cleaned up comments and console
* Unit tests for batchChanges
Much more scope for some crazy content
* Logging tweaks
- disable logging by default
- initWorker comes in too late to affect initial trace, so just rely on the `debug` to toggle at runtime
Co-authored-by: Richard Cox <richard.cox@suse.com>
- When there are over a configurable amount of resources to display in a list force the user to select a single namespace and use it to fetch resources related to the list
- Disabled by default, this can be enabled via the usual Global Settings --> Performance setting as usual
Functional Comments
- Gates for forcing the filter (count, resource type is namespaced, etc) apply only to the resources shown in the list.
- For example PV's aren't namespaced, so no enforced filtering. However they fetch PVC's which are namespaced
- For example we could have 10 resources to show in the list, but the resource types list component fetches 10000 other resources. The secondary resources are not taken in to account
- If we're under the threshold and have fetched all resources, if in that session we go over the threshold we won't fetch NS specific resources (because we have them all already)
- If we're over the threshold and have fetched namespaced resources, if in that session we go under the threshold we will fetch all resources
- If we're over the threshold and have fetched namespaced resources, going to a page that needs them all will result in us fetching them all (for instance from `events` to `cluster dashboard`)
- Deselecting a namespace and selecting it again should not kick off another http request
General Commit Comments
- The threshold to enforce the filter is set at 1500 as per manual fresh and incremental loading
- Optimised some code in ResourceList, resource-fetch and $loadingResources
* Prevent ESLint to format generated paths
* Add HTML ESLint Rules
* Format all the Vue HTML templates except Harvester
* Exclude Harvester package from linting HTML, keeping old formatting rules
- update preferences layout
- disable interval by setting, rather than 0
- move out preference update logic to gc land
- ensure new interval time period picked up by always stopping interval
- Don't cache gc state in store (avoids any chance of prompting change detection when update type accessed time)
- use TS instead of JS where possible
- split out two prompts into own file (on interval and on route change)
Remove resources from the store if they meet certain criteria. This will reduce the memory footprint of the dashboard and load on the backend (less watchers for large collections)
- GC is disabled by default and can be enabled via the Global Settings --> Performance tab
- User can configure
- The age in milliseconds in which a resource has to exceed in order to be gc'd
- The count which a resource has to exceed in order to be gc'd
- GC occures in stores that have it enabled
- ATM this is just the `cluster` store... but could be enabled for dashboard-store's such as the harvester one (one liner plus optional `gcIgnoreTypes` override for ignoring types)
- GC will be kicked off in two cases
- Route Change for a logged in state
- At a given interval
- Resource type _not_ GC'd if
- The store is ignoring the type
- For example the `cluster` store doesn't want to gc things like `schema` and `count`
- We're going to a page for that resource (list, detail, etc)
- For example don't GC pods if we're going to a pods page
- The last time the resource was accessed was recently
- We store the resource accessed time via hooking into actions and getters
- Setting the last accessed time will cause watchers of that type to trigger (only an issue for duplicate watchers)... but importantly not watchers of other types
- The resource is being used in the current page/context
- We store the route changed time and compare it to the resource accessed time
- There's too few resources
- We might as well keep them to avoid a network request to re-populate
// TODO:
- Should additional features be added to preferences
- if GC on route change is enabled
- if GC on interval is enabled, and how often it runs
- Sensible default preferences
- Remove some logging
* add logic to login function to make sure incremental loading is enabled by default for all users who dont have the setting yet populated
* lower default threshold for performance settings to 1500
* simplify way we enable incremental loading by default when no setting is set
Co-authored-by: Alexandre Alves <aalves@Alexandres-MacBook-Pro.local>
- Following UX review this is the first set of changes
- The text on the settings page has been improved
- The background of the incremental loader has been removed
- Fixed a typo (`UI_PERFORNMANCE`)
- Add manual refresh experimental feature to specific list views
- Add incremental loading experimental feature to specific list views
Co-authored-by: Alexandre Alves <aalves@Alexandres-MBP.lan>
Co-authored-by: Alexandre Alves <aalves@Alexandres-MacBook-Pro.local>
Co-authored-by: Richard Cox <richard.cox@suse.com>
Richard Cox
Nancy
Cody Jackson
Richard Cox
Richard
Giuseppe Leo
Sorin
Sean-McQ
Alexandre Alves
Jordon Leach