Decouple xds capacity controller and autopilot
This prevents a potential bug where autopilot deadlocks while attempting
to execute `AutopilotDelegate.NotifyState()` on an xdscapacity controller
that stopped consuming messages.
* Panic when controllers attempt to make invalid requests to the resource service
This will help to catch bugs in tests that could cause infinite errors to be emitted.
* Disable the API GW v2 controller
With the previous commit, this would cause a server to panic due to watching a type which has not yet been created/registered.
* Ensure that a test server gets the full type registry instead of constructing its own
* Skip TestServer_ControllerDependencies
* Fix peering tests so that they use the full resource registry.
* NET-7630 - Fix TXT record creation on node queries
* NET-7631 - Fix Node records that point to external/ non-IP addresses
* NET-7630 - Fix TXT record creation on node queries
Fix issue with persisting proxy-defaults
This resolves an issue introduced in hashicorp/consul#19829
where the proxy-defaults configuration entry with an HTTP protocol
cannot be updated after it has been persisted once and a router
exists. This occurs because the protocol field is not properly
pre-computed before being passed into validation functions.
* DNS V2 - Revise discovery result to have service and node name and address fields.
* NET-7488 - dns v2 add support for prepared queries in catalog v1 data model (#20470)
NET-7488 - dns v2 add support for prepared queries in catalog v1 data model.
The new controller caches are initialized before the DependencyMappers or the
Reconciler run, but importantly they are not populated. The expectation is that
when the WatchList call is made to the resource service it will send an initial
snapshot of all resources matching a single type, and then perpetually send
UPSERT/DELETE events afterward. This initial snapshot will cycle through the
caching layer and will catch it up to reflect the stored data.
Critically the dependency mappers and reconcilers will race against the restoration
of the caches on server startup or leader election. During this time it is possible a
mapper or reconciler will use the cache to lookup a specific relationship and
not find it. That very same reconciler may choose to then recompute some
persisted resource and in effect rewind it to a prior computed state.
Change
- Since we are updating the behavior of the WatchList RPC, it was aligned to
match that of pbsubscribe and pbpeerstream using a protobuf oneof instead of the enum+fields option.
- The WatchList rpc now has 3 alternating response events: Upsert, Delete,
EndOfSnapshot. When set the initial batch of "snapshot" Upserts sent on a new
watch, those operations will be followed by an EndOfSnapshot event before beginning
the never-ending sequence of Upsert/Delete events.
- Within the Controller startup code we will launch N+1 goroutines to execute WatchList
queries for the watched types. The UPSERTs will be applied to the nascent cache
only (no mappers will execute).
- Upon witnessing the END operation, those goroutines will terminate.
- When all cache priming routines complete, then the normal set of N+1 long lived
watch routines will launch to officially witness all events in the system using the
primed cached.
* Trigger the v1 compat exported-services controller when the v1 config entry is modified.
* Hook up exported-services config entries to the event publisher.
* Add tests to the v2 exported services shim.
* Use the local materializer trigger updates on the v1 compat exported services controller when exported-services config entries are modified.
* stop sleeping when context is cancelled
* Add Stop method to telemetry provider
Stop the main loop of the provider and set the config
to disabled.
* Add interface for telemetry provider
Added for easier testing. Also renamed Run to Start, which better
fits with Stop.
* Add Stop method to HCP manager
* Add manager interface, rename implementation
Add interface for easier testing, rename existing Manager to HCPManager.
* Stop HCP manager in link Finalizer
* Attempt to cleanup if resource has been deleted
The link should be cleaned up by the finalizer, but there's an edge
case in a multi-server setup where the link is fully deleted on one
server before the other server reconciles. This will cover the case
where the reconcile happens after the resource is deleted.
* Add a delete mananagement token function
Passes a function to the HCP manager that deletes the management token
that was initially created by the manager.
* Delete token as part of stopping the manager
* Lock around disabling config, remove descriptions
* Check for ACL write permissions on write
Link eventually will be creating a token, so require acl:write.
* Convert Run to Start, only allow to start once
* Always initialize HCP components at startup
* Support for updating config and client
* Pass HCP manager to controller
* Start HCP manager in link resource
Start as part of link creation rather than always starting. Update
the HCP manager with values from the link before starting as well.
* Fix metrics sink leaked goroutine
* Remove the hardcoded disabled hostname prefix
The HCP metrics sink will always be enabled, so the length of sinks will
always be greater than zero. This also means that we will also always
default to prefixing metrics with the hostname, which is what our
documentation states is the expected behavior anyway.
* Add changelog
* Check and set running status in one method
* Check for primary datacenter, add back test
* Clarify merge reasoning, fix timing issue in test
* Add comment about controller placement
* Expand on breaking change, fix typo in changelog
* If a workload does not implement a port, it should not be included in the list of endpoints for the Envoy cluster for that port.
* Adds tenancy tests for xds controller and xdsv2 resource generation, and adds all those files.
* The original change in this PR was for filtering the list of endpoints by the port being routed to (bullet 1). Since I made changes to sidecarproxycontroller golden files, I realized some of the golden files were unused because of the tenancy changes, so when I deleted those, that broke xds controller tests which weren't correctly using tenancy. So when I fixed that, then the xdsv2 tests broke, so I added tenancy support there too. So now, from sidecarproxy controller -> xds controller -> xdsv2 we now have tenancy support and all the golden files are lined up.
* API Gateway proto
* fix lint issue
* new line
* run make proto format
* checkpoint
* stub
* Update internal/mesh/internal/controllers/apigateways/controller.go
* Change logging of registered v2 resource endpoints to add /api prefix
Previous:
agent.http: Registered resource endpoint: endpoint=/demo/v1/executive
New:
agent.http: Registered resource endpoint: endpoint=/api/demo/v1/executive
This reduces confusion when attempting to call the APIs after looking at
the logs.
* Move config-dependent methods to separate package
In order to reuse the fetching and file creation part of the
bootstrap package, move the code that would cause cyclical
dependencies to a different package.
* Export needed bootstrap methods and variables
Also add back validating persisted config and update tests.
* Add support to check for just management token
Add a new method that fetches the bootstrap configuration only if
there isn't a valid management token file instead of checking for
all the hcp-config files.
* Pass data dir as a dependency to link controller
The link controller needs to check the data directory for
the hcp-config files.
* Fetch bootstrap config for token in controller
Load the management token when reconciling a link resource, which will
fetch the agent boostrap configuration if the token is not already
persisted locally. Skip this step if the cluster is in read-only mode.
* Validate resource ID format in link creation
* Handle unauthorized and forbidden errors
Check for 401 and 403s when making GNM requests, exit bootstrap fetch
loop and return specific failure statuses for link.
* Move test function to a testing file
* Log load and status write errors
* Exported services api implemented
* Tests added, refactored code
* Adding server tests
* changelog added
* Proto gen added
* Adding codegen changes
* changing url, response object
* Fixing lint error by having namespace and partition directly
* Tests changes
* refactoring tests
* Simplified uniqueness logic for exported services, sorted the response in order of service name
* Fix lint errors, refactored code
Add case insensitive param on service route match
This commit adds in a new feature that allows service routers to specify that
paths and path prefixes should ignore upper / lower casing when matching URLs.
Co-authored-by: Derek Menteer <105233703+hashi-derek@users.noreply.github.com>
Ultimately we will have to rectify wan federation with v2 catalog adjacent
experiments, but for now blanket prevent usage of the resource-apis,
v2dns, and v2tenancy experiments in secondary datacenters.
* Create HCP management token in HCP manager
* Change InitializeManagementToken to ManagementTokenUpserter
* Implement and use management token upsert function
* Fix race condition in test
* Add idea for improvement as comment
* Return early in upsertManagementToken if token exists
* Add Initializer to the controller
The Initializer adds support for running any required initialization
steps when the controller is first started.
* Implement HCP Link initializer
The link initializer will create a Link resource if the
cloud configuration has been set.
* Simplify retry logic and testing
* Remove internal retry, replace with logging logic
This add a fix to properly verify the gateway mode before creating a watch specific to mesh gateways. This watch have a high performance cost and when mesh gateways are not used is not used.
This also adds an optimization to only return the nodes when watching the Internal.ServiceDump RPC to avoid unnecessary disco chain compilation. As watches in proxy config only need the nodes.
* Option to set HCP client at runtime
Allows us to initially set a nil HCP client for the
telemetry provider and update it later.
* Set telemetry provider HCP client in HCP manager
Set the telemetry provider as a dependency and pass it to
the manager. Update the telemetry provider's HCP client
when the HCP manager starts.
* Add a provider interface for the metrics client
This provider will allow us to configure and reconfigure the
retryable HTTP client and the headers for the metrics client.
* Move HTTP retryable client to separate file
Copied directly from the metrics client.
* Abstract HCP specific values in HTTP client
Remove HCP specific references and instead initiate with
a generic TLS configuration and authentication source.
* Set up HTTP client and headers in the provider
Move setup from the metrics client to the HCP telemetry
provider.
* Update the telemetry provider in the HCP manager
Initialize the provider without the HCP configs and then update
it in the HCP manager to enable it.
* Improve test assertion, fix method comment
* Move client provider to metrics client
* Stop the manager on setup error
* Add separate lock for http configuration
* Start telemetry provider in HCP manager
* Update HCP client and config as part of Run
* Remove option to set config at initialization
* Simplify and clean up setting HCP configs
* Add test for telemetry provider Run method
* Fix race condition
* Use clone of HTTP headers
* Only allow initial update and run once
* Increase timeouts for flakey peering test.
* Various test fixes.
* Fix race condition in reconcilePeering.
This resolves an issue where a peering object in the state store was
incorrectly mutated by a function, resulting in the test being flagged as
failing when the -race flag was used.
To fix an issue displaying the current reloaded config in the
v1/agent/self endpoint #18681 caused the agent's internal
config struct member to be deepcopied and replaced on reload.
This is not safe because the field is not protected by a lock, nor
should it be due to how it is accessed by the rest of the system.
This PR does the same deepcopy, but into a new field solely for
the point of capturing the current reloaded values for display
purposes. If there has been no reload then the original config is used.
* Implement In-Process gRPC for use by controller caching/indexing
This replaces the pipe base listener implementation we were previously using. The new style CAN avoid cloning resources which our controller caching/indexing is taking advantage of to not duplicate resource objects in memory.
To maintain safety for controllers and for them to be able to modify data they get back from the cache and the resource service, the client they are presented in their runtime will be wrapped with an autogenerated client which clones request and response messages as they pass through the client.
Another sizable change in this PR is to consolidate how server specific gRPC services get registered and managed. Before this was in a bunch of different methods and it was difficult to track down how gRPC services were registered. Now its all in one place.
* Fix race in tests
* Ensure the resource service is registered to the multiplexed handler for forwarding from client agents
* Expose peer streaming on the internal handler
* Add HCCLink resource type
* Register HCCLink resource type with basic validation
* Add validation for required fields
* Add test for default ACLs
* Add no-op controller for HCCLink
* Add resource-apis semantic validation check in hcclink controller
* Add copyright headers
* Rename HCCLink to Link
* Add hcp_cluster_url to link proto
* Update 'disabled' reason with more detail
* Update link status name to consul.io/hcp/link
* Change link version from v1 to v2
* Use feature flag/experiment to enable v2 resources with HCP
* Update SCADA provider version
Also update mocks for SCADA provider.
* Create SCADA provider w/o HCP config, then update
Adds a placeholder config option to allow us to initialize a SCADA provider
without the HCP configuration. Also adds an update method to then add the
HCP configuration. We need this to be able to eventually always register a
SCADA listener at startup before the HCP config values are known.
* Pass cloud configuration to HCP manager
Save the entire cloud configuration and pass it to the HCP
manager.
* Update and start SCADA provider in HCP manager
Move config updating and starting to the HCP manager. The HCP manager
will eventually be responsible for all processes that contribute
to linking to HCP.
* NET-6945 - Replace usage of deprecated Envoy field envoy.config.core.v3.HeaderValueOption.append
* update proto for v2 and then update xds v2 logic
* add changelog
* Update 20078.txt to be consistent with existing changelog entries
* swap enum values tomatch envoy.
* NET-6946 - Replace usage of deprecated Envoy field envoy.config.route.v3.HeaderMatcher.safe_regex_match
* removing unrelated changes
* update golden files
* do not set engine type
* NET-6426 Create ProxyStateTemplate when reconciling MeshGateway resource
* Add TODO for switching fetch method based on gateway type
* Use gateway-kind in workload metadata instead of owner reference
* Create ProxyStateTemplate builder for gatewayproxy controller
* Update to use new controller interface
* Add copyright headers
* Set correct name for ProxyStateTemplate identity reference
* Generate empty ProxyStateTemplate by fetching MeshGateway
This cheats and looks up the MeshGateway directly. In the future, we will need a Workload => xGateway mapper
* Specify owner reference when writing ProxyStateTemplate
* Update dependency mapper to account for multiple controllers per resource type
* Regenerate v2 resource dependencies map
* Add helpful trace logs, tag TODOs with ticket identifiers
This commit fixes an issue where the partition was not properly set
on the peering query failover target created from sameness-groups.
Before this change, it was always empty, meaning that the data
would be queried with respect to the default partition always. This
resulted in a situation where a PQ that was attempting to use a
sameness-group for failover would select peers from the default
partition, rather than the partition of the sameness-group itself.
* add a hash to config entries when normalizing
* add GetHash and implement comparing hashes
* only update if the Hash is different
* only update if the Hash is different and not 0
* fix proto to include the Hash
* fix proto gen
* buf format
* add SetHash and fix tests
* fix config load tests
* fix state test and config test
* recalculate hash when restoring config entries
* fix snapshot restore test
* add changelog
* fix missing normalize, fix proto indexes and add normalize test
When a large number of upstreams are configured on a single envoy
proxy, there was a chance that it would timeout when waiting for
ClusterLoadAssignments. While this doesn't always immediately cause
issues, consul-dataplane instances appear to consistently drop
endpoints from their configurations after an xDS connection is
re-established (the server dies, random disconnect, etc).
This commit adds an `xds_fetch_timeout_ms` config to service registrations
so that users can set the value higher for large instances that have
many upstreams. The timeout can be disabled by setting a value of `0`.
This configuration was introduced to reduce the risk of causing a
breaking change for users if there is ever a scenario where endpoints
would never be received. Rather than just always blocking indefinitely
or for a significantly longer period of time, this config will affect
only the service instance associated with it.
This fixes the following race condition:
- Send update endpoints
- Send update cluster
- Recv ACK endpoints
- Recv ACK cluster
Prior to this fix, it would have resulted in the endpoints NOT existing in
Envoy. This occurred because the cluster update implicitly clears the endpoints
in Envoy, but we would never re-send the endpoint data to compensate for the
loss, because we would incorrectly ACK the invalid old endpoint hash. Since the
endpoint's hash did not actually change, they would not be resent.
The fix for this is to effectively clear out the invalid pending ACKs for child
resources whenever the parent changes. This ensures that we do not store the
child's hash as accepted when the race occurs.
An escape-hatch environment variable `XDS_PROTOCOL_LEGACY_CHILD_RESEND` was
added so that users can revert back to the old legacy behavior in the event
that this produces unknown side-effects. Visit the following thread for some
extra context on why certainty around these race conditions is difficult:
https://github.com/envoyproxy/envoy/issues/13009
This bug report and fix was mostly implemented by @ksmiley with some minor
tweaks.
Co-authored-by: Keith Smiley <ksmiley@salesforce.com>
* Add CE version of gateway-upstream-disambiguation
* Use NamespaceOrDefault and PartitionOrDefault
* Add Changelog entry
* Remove the unneeded reassignment
* Use c.ID()
The client.rpc metric now excludes internal retries for consistency
with client.rpc.exceeded and client.rpc.failed. All of these metrics
now increment at most once per RPC method call, allowing for
accurate calculation of failure / rate limit application occurrence.
Additionally, if an RPC fails because no servers are present,
client.rpc.failed is now incremented.
* Add a make target to run lint-consul-retry on all the modules
* Cleanup sdk/testutil/retry
* Fix a bunch of retry.Run* usage to not use the outer testing.T
* Fix some more recent retry lint issues and pin to v1.4.0 of lint-consul-retry
* Fix codegen copywrite lint issues
* Don’t perform cleanup after each retry attempt by default.
* Use the common testutil.TestingTB interface in test-integ/tenancy
* Fix retry tests
* Update otel access logging extension test to perform requests within the retry block
Dan Stough