## HTTPAdapter (#5637)
## Ember upgrade 2.18 > 3.12 (#6448)
### Proxies can no longer get away with not calling _super
This means that we can't use create anymore to define dynamic methods.
Therefore we dynamically make 2 extended Proxies on demand, and then
create from those. Therefore we can call _super in the init method of
the extended Proxies.
### We aren't allowed to reset a service anymore
We never actually need to now anyway, this is a remnant of the refactor
from browser based confirmations. We fix it as simply as possible here
but will revisit and remove the old browser confirm functionality at a
later date
### Revert classes to use ES5 style to workaround babel transp. probs
Using a mixture of ES6 classes (and hence super) and arrow functions
means that when babel transpiles the arrow functions down to ES5, a
reference to this is moved before the call to super, hence causing a js
error.
Furthermore, we the testing environment no longer lets use use
apply/call on the constructor.
These errors only manifests during testing (only in the testing
environment), the application itself runs fine with no problems without
this change.
Using ES5 style class definitions give us freedom to do all of the above
without causing any errors, so we reverted these classes back to ES5
class definitions
### Skip test that seems to have changed due to a change in RSVP timing
This test tests a usecase/area of the API that will probably never ever
be used, it was more testing out the API. We've skipped the test for now
as this doesn't affect the application itself, but left a note to come
back here later to investigate further
### Remove enumerableContentDidChange
Initial testing looks like we don't need to call this function anymore,
the function no longer exists
### Rework Changeset.isSaving to take into account new ember APIs
Setting/hanging a computedProperty of an instantiated object no longer
works. Move to setting it on the prototype/class definition instead
### Change how we detect whether something requires listening
New ember API's have changed how you can detect whether something is a
computedProperty or not. It's not immediately clear if its even possible
now. Therefore we change how we detect whether something should be
listened to or not by just looking for presence of `addEventListener`
### Potentially temporary change of ci test scripts to ensure deps exist
All our tooling scripts run through a Makefile (for people familiar with
only using those), which then call yarn scripts which can be called
independently (for people familar with only using yarn).
The Makefile targets always check to make sure all the dependencies are
installed before running anything that requires them (building, testing
etc).
The CI scripts/targets didn't follow this same route and called the yarn
scripts directly (usually CI builds a cache of the dependencies first).
For some reason this cache isn't doing what it usually does, and it
looks as though, in CI, ember isn't installed.
This commit makes the CI scripts consistently use the same method as all
of the other tooling scripts (Makefile target > Install Deps if
required > call yarn script). This should install the dependencies if
for some reason the CI cache building doesn't complete/isn't successful.
Potentially this commit may be reverted if, the root of the problem is
elsewhere, although consistency is always good, so it might be a good
idea to leave this commit as is even if we need to debug and fix things
elsewhere.
### Make test-parallel consistent with the rest of the tooling scripts
As we are here making changes for CI purposes (making test-ci
consistent), we spotted that test-parallel is also inconsistent and also
the README manual instructions won't work without `ember` installed
globally.
This commit makes everything consistent and changes the manual
instructions to use the local ember instance that gets installed via
yarn
### Re-wrangle catchable to fit with new ember 3.12 APIs
In the upgrade from ember 3.8 > 3.12 the public interfaces for
ComputedProperties have changed slightly. `meta` is no longer a public
property of ComputedProperty but of a ComputedDecoratorImpl mixin
instead.
7e4ba1096e/packages/%40ember/-internals/metal/lib/computed.ts (L725)
There seems to be no way, by just using publically available
methods, to replicate this behaviour so that we can create our own
'ComputedProperty` factory via injecting the ComputedProperty class as
we did previously.
3f333bada1/ui-v2/app/utils/computed/factory.js (L1-L18)
Instead we dynamically hang our `Catchable` `catch` method off the
instantiated ComputedProperty. In doing it like this `ComputedProperty`
has already has its `meta` method mixed in so we don't have to manually
mix it in ourselves (which doesn't seem possible)
This functionality is only used during our work in trying to ensure
our EventSource/BlockingQuery work was as 'ember-like' as possible (i.e.
using the traditional Route.model hooks and ember-like Controller
properties). Our ongoing/upcoming work on a componentized approach to
data a.k.a `<DataSource />` means we will be able to remove the majority
of the code involved here now that it seems to be under an amount of
flux in ember.
### Build bindata_assetfs.go with new UI changes
- yarn upgrade consul-api-double which includes `status/leader`
- add all the ember-data things required to call a new endpoint
- Pass the new leader variable through to the template
- use the new leader variable in the template to set a leader
- add acceptance testing to verify leaders are highlighted
- Change testing navigation/api requests to status/leader (on the node listing page, status/leader is now the last get request to
be called).
- Template whitespace commit (less indenting)
- adds a test to to assert no errors happen with an unelected leader
-Enable blocking queries by default
-Change assertion to check for the last PUT request, not just any request for session destruction from a node page.
Since we've now turned on blocking queries by default this means that a
second GET request is made after the PUT request that we are asserting
for but before the assertion itself, this meant the assertion failed. We
double checked this by turning off blocking queries for this test using
```
And settings from yaml
---
consul:client:
blocking: 0
---
```
which made the test pass again.
As moving forwards blocking queries will be on by default, we didn't
want to disable blocking queries for this test, so we now assert the
last PUT request specifically. This means we continue to assert that the
session has been destroyed but means we don't get into problems of
ordering of requests here
- Removes 'type' icons (basically the proxy icon, not the text itself)
- Add support for Mesh Gateways plus their addresses
This adds a 'Mesh Gateway' type label to service and service instance
pages, plus a new 'Addresses' tab if the service is a Mesh Gateway
showing a table of addresses for the service - plus tests
* ui: Normal proxies line to services, sidecars to instances
Following on from https://github.com/hashicorp/consul/pull/5933 we
noticed that 'normal' proxies should link to the service, rather than
the service instance. Additionally proxy 'searching' within the
repository should take into account the name of the node that the
originating service is on (sidecar proxies are generally co-located)
Added an additional test here to prove that a sidecar-proxy with the
same service id but on a different node does not show the sidecar proxy
link.
1. All {{ivy-codemirror}} components need 'refreshing' when they become
visible via our own `didAppear` method on the `{{code-editor}}`
component
(also see:)
- https://github.com/hashicorp/consul/pull/4190#discussion_r193270223
- 73db111db8 (r225264296)
2. On initial investigation, it looks like the component we are using
for the code editor doesn't distinguish between setting its `value`
programatically and a `keyup` event, i.e. an interaction from the user.
We currently pretend that whenever its `value` changes, it is a `keyup`
event. This means that when we reset the `value` to `""`
programmatically for form resetting purposes, a 'pretend keyup' event
would also be fired, which would in turn kick off the validation, which
would fail and show an error message for empty values in other fields of
the form - something that is perfectly valid if you haven't typed
anything yet. We solved this by checking for `isPristine` on fields that
are allowed to be empty before you have typed anything.
Previously we were creating a fake event and amending the name of the
fake event, this meant that other `event.target` properties weren't
being passed through (in this instance `checked`) this changes the
approach to not use fake events, and allows you to overwrite the name
that the form uses for `handleEvent`
1. Includes Datacenter variable for intperolation
2. Amends text on the Settings page to reflect new keyword
3. Adds further acceptance testing around the new dashboard buttons
Adds support for ACL Roles and Service Identities CRUD, along with necessary changes to Tokens, and the CSS improvements required.
Also includes refinements/improvements for easier testing of deeply nested components.
1. ember-data adapter/serializer/model triplet for Roles
2. repository, form/validations and searching filter for Roles
3. Moves potentially, repeated, or soon to to repeated functionality
into a mixin (mainly for 'many policy' relationships)
4. A few styling tweaks for little edge cases around roles
5. Router additions, Route, Controller and templates for Roles
Also see:
* UI: ACL Roles cont. plus Service Identities (#5661 and #5720)
If a service instance show page is being viewed and the service instance
is deregistered, this closes the blocking query for the proxy as well as
the instance (the instances query will be clsed on the error)
Also adds skipped tests to nag in future
Previously the tomography wasn't using ember `get` so proxy updates
(specifically here whilst receiving a blocking update) wasn't working.
This adds `get` here until we update to newer `get`less ember and also
refactors slightly removing `n` and using `distance.length` instead
Skipped tests are adding here to nag us to come back here at some point.
This commit includes several pieces of functionality to enable services
to be removed and the page to present information that this has happened
but also keep the deleted information on the page. Along with the more
usual blocking query based listing.
To enable this:
1. Implements `meta` on the model (only available on collections in
ember)
2. Adds new `catchable` ComputedProperty alongside a `listen` helper for
working with specific errors that can be thrown from EventSources in an
ember-like way. Briefly, normal computed properties update when a
property changes, EventSources can additionally throw errors so we can
catch them and show different visuals based on that.
Also:
Add support for blocking queries on the service instance detail page
1. Previous we could return undefined when a service instance has no
proxy, but this means we have nothing to attach `meta` to. We've changed
this to return an almost empty object, so with only a meta property.
At first glance there doesn't seem to be any way to provide a proxy
object to templates and be able to detect whether it is actually null
or not so we instead change some conditional logic in the templates to
detect the property we are using to generate the anchor.
2. Made a `pauseUntil` test helper function for steps where we wait for
things. This helps for DRYness but also means if we can move away from
setInterval to something else later, we can do it in one place
3. Whilst running into point 1 here, we managed to make the blocking
queries eternally loop. Whilst this is due to an error in the code and
shouldn't ever happen whilst in actual use, we've added an extra check
so that we only recur/loop the blocking query if the previous response has a
`meta.cursor`
Adds support for blocking queries on the node detail page (#5489)
1. Moves data re-shaping for the templates variables into a repository
so they are easily covered by blocking queries (into coordinatesRepo)
2. The node API returns a 404 as signal for deregistration, we also
close the sessions and coordinates blocking queries when this happens
This commit includes several pieces of functionality to enable services
to be removed and the page to present information that this has happened
but also keep the deleted information on the page. Along with the more
usual blocking query based listing.
To enable this:
1. Implements `meta` on the model (only available on collections in
ember)
2. Adds new `catchable` ComputedProperty alongside a `listen` helper for
working with specific errors that can be thrown from EventSources in an
ember-like way. Briefly, normal computed properties update when a
property changes, EventSources can additionally throw errors so we can
catch them and show different visuals based on that.
More recommendations for blocking queries clients was added here:
https://github.com/hashicorp/consul/pull/5358
This commit mainly adds cursor/index validation/correction based on
these recommendations (plus tests)
The recommendations also suggest that clients should include rate
limiting. Because of this, we've moved the throttling out of Consul UI
specific code and into Blocking Query specific code. Currently the 'rate
limiting' in this commit only adds a sleep to every iteration of the
loop, which is not the recommended approach, but the code here organizes
the throttling functionality into something we can work with later to
provide something more apt.
* ui: Add forking based on service instance id existence
Proxies come in 2 flavours, 'normal' and sidecar. We know when a proxy
is a sidecar proxy based on whether a DestinationServiceID is set.
LocalServiceAddress and LocalServicePort are only relevant for sidecar
proxies.
This adds template logic to show different text depending on this
information.
Additionally adds test around connect proxies (#5418)
1. Adds page object for the instance detail page
2. Adds further scenario steps used in the tests
3. Adds acceptance testing around the instance detail page. Services
with proxies and the sidecar proxies and proxies themselves
4. Adds datacenter column for upstreams
5. Fixes bug routing bug for decision as to whether to request proxy
information or not
This gives more prominence to 'Service Instances' as opposed to 'Services'. It also begins to surface Connect related 'nouns' such as 'Proxies' and 'Upstreams' and begins to interconnect them giving more visibility to operators.
Various smaller changes:
1. Move healthcheck-status component to healthcheck-output
2. Create a new healthcheck-status component for showing the number of
checks plus its icon
3. Create a new healthcheck-info component to group multiple statuses
plus a different view if there are no checks
4. Componentize tag-list
- Maintain http headers as JSON-API meta for all API requests (#4946)
- Add EventSource ready for implementing blocking queries
- EventSource project implementation to enable blocking queries for service and node listings (#5267)
- Add setting to enable/disable blocking queries (#5352)
In 858b05fc31 (diff-46ef88aa04507fb9b039344277531584)
we removed encoding values in pathnames as we thought they were
eventually being encoded by `ember`. It looks like this isn't the case.
Turns out sometimes they are encoded sometimes they aren't. It's complicated.
If at all possible refer to the PR https://github.com/hashicorp/consul/pull/5206.
It's related to the difference between `dynamic` routes and `wildcard` routes.
Partly related to this is a decision on whether we urlencode the slashes within service names or not. Whilst historically we haven't done this, we feel its a good time to change this behaviour, so we'll also be changing services to use dynamic routes instead of wildcard routes. So service links will then look like /ui/dc-1/services/application%2Fservice rather than /ui/dc-1/services/application/service
Here, we define our routes in a declarative format (for the moment at least JSON) outside of Router.map, and loop through this within Router.map to set all our routes using the standard this.route method. We essentially configure our Router from the outside. As this configuration is now done declaratively outside of Router.map we can also make this data available to href-to and paramsFor, allowing us to detect wildcard routes and therefore apply urlencoding/decoding.
Where I mention 'conditionally' below, this is detection is what is used for the decision.
We conditionally add url encoding to the `{{href-to}}` helper/addon. The
reasoning here is, if we are asking for a 'href/url' then whatever we
receive back should always be urlencoded. We've done this by reusing as much
code from the original `ember-href-to` addon as possible, after this
change every call to the `{{href-to}}` helper will be urlencoded.
As all links using `{{href-to}}` are now properly urlencoded. We also
need to decode them in the correct place 'on the other end', so..
We also override the default `Route.paramsFor` method to conditionally decode all
params before passing them to the `Route.model` hook.
Lastly (the revert), as we almost consistently use url params to
construct API calls, we make sure we re-encode any slugs that have been
passed in by the user/developer. The original API for the `createURL`
function was to allow you to pass values that didn't need encoding,
values that **did** need encoding, followed by query params (which again
require url encoding)
All in all this should make the entire ember app url encode/decode safe.
In order to continue supporting the legacy ACL system, we replace
the 500 error from a non-existent `self` endpoint with a response of a
`null` `AccessorID` - which makes sense (a null AccessorID means old
API)
We then redirect the user to the old ACL pages which then gives a 403
if their token was wrong which then redirects them back to the login page.
Due to the multiple redirects and not wanting to test the validity of the token
before redirecting (thus calling the same API endpoint twice), it is not
straightforwards to turn the 'faked' response from the `self` endpoint
into an error (flash messages are 'lost' through multiple redirects).
In order to make this a slightly better experience, you can now return a
`false` during execution of an action requiring success/failure
feedback, this essentially skips the notification, so if the action is
'successful' but you don't want to show the notification, you can. This
resolves showing a successful notification when the `self` endpoint
response is faked. The last part of the puzzle is to make sure that the
global 403 catching error in the application Route also produces an
erroneous notification.
Please note this can only happen with a ui client using the new ACL
system when communicating with a cluster using the old ACL system, and
only when you enter the wrong token.
Lastly, further acceptance tests have been added around this
This commit also adds functionality to avoid any possible double
notification messages, to avoid UI overlapping
In some circumstances a consul 1.4 client could be running in an
un-upgraded 1.3 or lower cluster. Currently this gives a 500 error on
the new ACL token endpoint. Here we catch this specific 500 error/message
and set the users AccessorID to null. Elsewhere in the frontend we use
this fact (AccessorID being null) to decide whether to present the
legacy or the new ACL UI to the user.
Also:
- Re-adds in most of the old style ACL acceptance tests, now that we are keeping the old style UI
- Restricts code editors to HCL only mode for all `Rules` editing (legacy/'half legacy'/new style)
- Adds a [Stop using] button to the old style ACL rows so its possible to logout.
- Updates copy and documentation links for the upgrade notices
Having the code editor on removes the text area from the DOM, making it
more difficult to enter text in the text editor during testing. This
turns the code editor off whilst making edits during testing.
No changes to UI code
The mocks where using randomly generated `ExternalSources` this change
makes sure they are fixed so we can reliably test the values. No change
to actual UI code
The error notification was being shown on creation of an intention. This
was as a result of #4572 and/or #4572 and has not been included in a
release.
This includes a fix, plus tests to try to prevent any further regression.
1. Addition of external source icons for services marked as such.
2. New %with-tooltip css component (wip)
3. New 'no healthcheck' icon as external sources might not have
healthchecks, also minus icon on node cards in the service detail view
4. If a service doesn't have healthchecks, we use the [Services] tabs as the
default instead of the [Health Checks] tab in the Service detail page.
5. `css-var` helper. The idea here is that it will eventually be
replaced with pure css custom properties instead of having to use JS. It
would be nice to be able to build the css variables into the JS at build
time (you'd probably still want to specify in config which variables you
wanted available in JS), but that's possible future work.
Lastly there is probably a tiny bit more testing edits here than usual,
I noticed that there was an area where the dynamic mocking wasn't
happening, it was just using the mocks from consul-api-double, the mocks
I was 'dynamically' setting happened to be the same as the ones in
consul-api-double. I've fixed this here also but it wasn't effecting
anything until actually made certain values dynamic.
* Move notification texts to a slightly different layer (#4572)
* Further Simplify/refactor the actions/notification layer (#4573)
1. Move the 'with-feedback' actions to a 'with-blocking-action' mixin
which better describes what it does
2. Additional set of unit tests almost over the entire layer to prove
things work/add confidence for further changes
The multiple 'with-action' mixins used for every 'index/edit' combo are
now reduced down to only contain the functionality related to their
specific routes, i.e. where to redirect.
The actual functionality to block and carry out the action and then
notify are 'almost' split out so that their respective classes/objects do
one thing and one thing 'well'.
Mixins are chosen for the moment as the decoration approach used by
mixins feels better than multiple levels of inheritence, but I would
like to take this fuether in the future to a 'compositional' based
approach.
There is still possible further work to be done here, but I'm a lot
happier now this is reduced down into separate parts.
We now essentially do 2 redirects if you hit a `folder/`
1. If you visit `/ui/dc1/kv/folder/`, `consul` will redirect you to `/ui/dc1/kv/folder`
2. Once redirected to `/ui/dc1/kv/folder` via a 301, use ember/history
API to redirect you back to `/ui/dc1/kv/folder/`.
Bit long winded, but achieves what we want without having to get stuck
into `consul` itself to remove the 301 for the UI
John Cowen
John Cowen