In an upcoming change we will need to pass a grpc.ClientConnPool from
BaseDeps into Server. While looking at that change I noticed all of the
existing consulOption fields are already on BaseDeps.
Instead of duplicating the fields, we can create a struct used by
agent/consul, and use that struct in BaseDeps. This allows us to pass
along dependencies without translating them into different
representations.
I also looked at moving all of BaseDeps in agent/consul, however that
created some circular imports. Resolving those cycles wouldn't be too
bad (it was only an error in agent/consul being imported from
cache-types), however this change seems a little better by starting to
introduce some structure to BaseDeps.
This change is also a small step in reducing the scope of Agent.
Also remove some constants that were only used by tests, and move the
relevant comment to where the live configuration is set.
Removed some validation from NewServer and NewClient, as these are not
really runtime errors. They would be code errors, which will cause a
panic anyway, so no reason to handle them specially here.
secondaryIntermediateCertRenewalWatch was using `retryLoopBackoff` to
renew the intermediate certificate. Once it entered the inner loop and
started `retryLoopBackoff` it would never leave that.
`retryLoopBackoffAbortOnSuccess` will return when renewing is
successful, like it was intended originally.
The nodeCheck slice was being used as the first arg in append, which in some cases will modify the array backing the slice. This would lead to service checks for other services in the wrong event.
Also refactor some things to reduce the arguments to functions.
Creating a new readTxn does not work because it will not see the newly created objects that are about to be committed. Instead use the active write Txn.
Whenever an upsert/deletion of a config entry happens, within the open
state store transaction we speculatively test compile all discovery
chains that may be affected by the pending modification to verify that
the write would not create an erroneous scenario (such as splitting
traffic to a subset that did not exist).
If a single discovery chain evaluation references two config entries
with the same kind and name in different namespaces then sometimes the
upsert/deletion would be falsely rejected. It does not appear as though
this bug would've let invalid writes through to the state store so the
correction does not require a cleanup phase.
This commit refactors the state store usage code to track unique service
name changes on transaction commit. This means we only need to lookup
usage entries when reading the information, as opposed to iterating over
a large number of service indices.
- Take into account a service instance's name being changed
- Do not iterate through entire list of service instances, we only care
about whether there is 0, 1, or more than 1.
We add a WriteTxn interface for use in updating the usage memdb table,
with the forward-looking prospect of incrementally converting other
functions to accept interfaces.
As well, we use the ReadTxn in new usage code, and as a side effect
convert a couple of existing functions to use that interface as well.
Using the newly provided state store methods, we periodically emit usage
metrics from the servers.
We decided to emit these metrics from all servers, not just the leader,
because that means we do not have to care about leader election flapping
causing metrics turbulence, and it seems reasonable for each server to
emit its own view of the state, even if they should always converge
rapidly.
During gossip encryption key rotation it would be nice to be able to see if all nodes are using the same key. This PR adds another field to the json response from `GET v1/operator/keyring` which lists the primary keys in use per dc. That way an operator can tell when a key was successfully setup as primary key.
Based on https://github.com/hashicorp/serf/pull/611 to add primary key to list keyring output:
```json
[
{
"WAN": true,
"Datacenter": "dc2",
"Segment": "",
"Keys": {
"0OuM4oC3Os18OblWiBbZUaHA7Hk+tNs/6nhNYtaNduM=": 6,
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 6
},
"PrimaryKeys": {
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 6
},
"NumNodes": 6
},
{
"WAN": false,
"Datacenter": "dc2",
"Segment": "",
"Keys": {
"0OuM4oC3Os18OblWiBbZUaHA7Hk+tNs/6nhNYtaNduM=": 8,
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 8
},
"PrimaryKeys": {
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 8
},
"NumNodes": 8
},
{
"WAN": false,
"Datacenter": "dc1",
"Segment": "",
"Keys": {
"0OuM4oC3Os18OblWiBbZUaHA7Hk+tNs/6nhNYtaNduM=": 3,
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 8
},
"PrimaryKeys": {
"SINm887hKTzmMWeBNKTJReaTLX3mBEJKriDyt88Ad+g=": 8
},
"NumNodes": 8
}
]
```
I intentionally did not change the CLI output because I didn't find a good way of displaying this information. There are a couple of options that we could implement later:
* add a flag to show the primary keys
* add a flag to show json output
Fixes#3393.
Fixes#8466
Since Consul 1.8.0 there was a bug in how ingress gateway protocol
compatibility was enforced. At the point in time that an ingress-gateway
config entry was modified the discovery chain for each upstream was
checked to ensure the ingress gateway protocol matched. Unfortunately
future modifications of other config entries were not validated against
existing ingress-gateway definitions, such as:
1. create tcp ingress-gateway pointing to 'api' (ok)
2. create service-defaults for 'api' setting protocol=http (worked, but not ok)
3. create service-splitter or service-router for 'api' (worked, but caused an agent panic)
If you were to do these in a different order, it would fail without a
crash:
1. create service-defaults for 'api' setting protocol=http (ok)
2. create service-splitter or service-router for 'api' (ok)
3. create tcp ingress-gateway pointing to 'api' (fail with message about
protocol mismatch)
This PR introduces the missing validation. The two new behaviors are:
1. create tcp ingress-gateway pointing to 'api' (ok)
2. (NEW) create service-defaults for 'api' setting protocol=http ("ok" for back compat)
3. (NEW) create service-splitter or service-router for 'api' (fail with
message about protocol mismatch)
In consideration for any existing users that may be inadvertently be
falling into item (2) above, that is now officiall a valid configuration
to be in. For anyone falling into item (3) above while you cannot use
the API to manufacture that scenario anymore, anyone that has old (now
bad) data will still be able to have the agent use them just enough to
generate a new agent/proxycfg error message rather than a panic.
Unfortunately we just don't have enough information to properly fix the
config entries.
* changes some functions to return data instead of modifying pointer
arguments
* renames globalRPC() to keyringRPCs() to make its purpose more clear
* restructures KeyringOperation() to make it more understandable
NotifyShutdown was only used for testing. Now that t.Cleanup exists, we
can use that instead of attaching cleanup to the Server shutdown.
The Autopilot test which used NotifyShutdown doesn't need this
notification because Shutdown is synchronous. Waiting for the function
to return is equivalent.
Ensure that enabling AutoConfig sets the tls configurator properly
This also refactors the TLS configurator a bit so the naming doesn’t imply only AutoEncrypt as the source of the automatically setup TLS cert info.
Most of the groundwork was laid in previous PRs between adding the cert-monitor package to extracting the logic of signing certificates out of the connect_ca_endpoint.go code and into a method on the server.
This also refactors the auto-config package a bit to split things out into multiple files.
Replaces #7559
Running tests in parallel, with background goroutines, results in test output not being associated with the correct test. `go test` does not make any guarantees about output from goroutines being attributed to the correct test case.
Attaching log output from background goroutines also cause data races. If the goroutine outlives the test, it will race with the test being marked done. Previously this was noticed as a panic when logging, but with the race detector enabled it is shown as a data race.
The previous solution did not address the problem of correct test attribution because test output could still be hidden when it was associated with a test that did not fail. You would have to look at all of the log output to find the relevant lines. It also made debugging test failures more difficult because each log line was very long.
This commit attempts a new approach. Instead of printing all the logs, only print when a test fails. This should work well when there are a small number of failures, but may not work well when there are many test failures at the same time. In those cases the failures are unlikely a result of a specific test, and the log output is likely less useful.
All of the logs are printed from the test goroutine, so they should be associated with the correct test.
Also removes some test helpers that were not used, or only had a single caller. Packages which expose many functions with similar names can be difficult to use correctly.
Related:
https://github.com/golang/go/issues/38458 (may be fixed in go1.15)
https://github.com/golang/go/issues/38382#issuecomment-612940030
The fallback method would still work but it would get into a state where it would let the certificate expire for 10s before getting a new one. And the new one used the less secure RPC endpoint.
This is also a pretty large refactoring of the auto encrypt code. I was going to write some tests around the certificate monitoring but it was going to be impossible to get a TestAgent configured in such a way that I could write a test that ran in less than an hour or two to exercise the functionality.
Moving the certificate monitoring into its own package will allow for dependency injection and in particular mocking the cache types to control how it hands back certificates and how long those certificates should live. This will allow for exercising the main loop more than would be possible with it coupled so tightly with the Agent.
When the race detector is enabled we see this test fail occasionally. The reordering of execution seems to make it possible for the snapshot splice to happen before any events are published to the topicBuffers.
We can handle this case in the test the same way it is handled by a subscription, by proceeding to the next event.
This change was mostly automated with the following
First generate a list of functions with:
git grep -o 'Store) \([^(]\+\)(tx \*txn' ./agent/consul/state | awk '{print $2}' | grep -o '^[^(]\+'
Then the list was curated a bit with trial/error to remove and add funcs
as necessary.
Finally the replacement was done with:
dir=agent/consul/state
file=${1-funcnames}
while read fn; do
echo "$fn"
sed -i -e "s/(s \*Store) $fn(/$fn(/" $dir/*.go
sed -i -e "s/s\.$fn(/$fn(/" $dir/*.go
sed -i -e "s/s\.store\.$fn(/$fn(/" $dir/*.go
done < $file
Making these functions allows them to be used without introducing
an artificial dependency on the struct. Many of these will be called
from streaming Event processors, which do not have a store.
This change is being made ahead of the streaming work to get to reduce
the size of the streaming diff.
Move the subscription context to Next. context.Context should generally
never be stored in a struct because it makes that struct only valid
while the context is valid. This is rarely obvious from the caller.
Adds a forceClosed channel in place of the old context, and uses the new
context as a way for the caller to stop the Subscription blocking.
Remove some recursion out of bufferImte.Next. The caller is already looping so we can continue
in that loop instead of recursing. This ensures currentItem is updated immediately (which probably
does not matter in practice), and also removes the chance that we overflow the stack.
NextNoBlock and FollowAfter do not need to handle bufferItem.Err, the caller already
handles it.
Moves filter to a method to simplify Next, and more explicitly separate filtering from looping.
Also improve some godoc
Only unwrap itemBuffer.Err when necessary
EventPublisher was receiving TopicHandlers, which had a couple of
problems:
- ChangeProcessors were being grouped by Topic, but they completely
ignored the topic and were performed on every change
- ChangeProcessors required EventPublisher to be aware of database
changes
By moving ChangeProcesors out of EventPublisher, and having Publish
accept events instead of changes, EventPublisher no longer needs to
be aware of these things.
Handlers is now only SnapshotHandlers, which are still mapped by Topic.
Also allows us to remove the small 'db' package that had only two types.
They can now be unexported types in state.
The EventPublisher is the central hub of the PubSub system. It is toughly coupled with much of
stream. Some stream internals were exported exclusively for EventPublisher.
The two Subscribe cases (with or without index) were also awkwardly split between two packages. By
moving EventPublisher into stream they are now both in the same package (although still in different files).
Also store the index in Changes instead of the Txn.
This change is in preparation for movinng EventPublisher to the stream package, and
making handleACLUpdates async once again.
It is critical that Unsubscribe be called with the same pointer to a
SubscriptionRequest that was used to create the Subscription. The
docstring made that clear, but it sill allowed a caler to get it wrong by
creating a new SubscriptionRequest.
By hiding this detail from the caller, and only exposing an Unsubscribe
method, it should be impossible to fail to Unsubscribe.
Also update some godoc strings.
Use a separate lock for subscriptions.ByToken to allow it to happen synchronously
in the commit flow.
This removes the need to create a new txn for the goroutine, and removes
the need for EventPublisher to contain a reference to DB.
Many of the fields are only needed in one place, and by using a closure
they can be removed from the struct. This reduces the scope of the variables
making it esier to see how they are used.
Otherwise the test will run with exactly the same values each time.
By printing the seed we can attempt to reproduce the test by adding an env var to override the seed
Make topicRegistry use functions instead of unbound methods
Use a regular memDB in EventPublisher to remove a reference cycle
Removes the need for EventPublisher to use a store
Also remove secretHash, which was used to hash tokens. We don't expose
these tokens anywhere, so we can use the string itself instead of a
Hash.
Fix acl_events_test.go for storing a structs type.
This is instead of having the AutoConfigBackend interface provide functions for retrieving them.
NOTE: the config is not reloadable. For now this is fine as we don’t look at any reloadable fields. If that changes then we should provide a way to make it reloadable.
In all cases (oss/ent, client/server) this method was returning a value from config. Since the
value is consistent, it doesn't need to be part of the delegate interface.
A query made with AllowNotModifiedResponse and a MinIndex, where the
result has the same Index as MinIndex, will return an empty response
with QueryMeta.NotModified set to true.
Co-authored-by: Pierre Souchay <pierresouchay@users.noreply.github.com>
The initial auto encrypt CSR wasn’t containing the user supplied IP and DNS SANs. This fixes that. Also We were configuring a default :: IP SAN. This should be ::1 instead and was fixed.
Highlights:
- add new endpoint to query for intentions by exact match
- using this endpoint from the CLI instead of the dump+filter approach
- enforcing that OSS can only read/write intentions with a SourceNS or
DestinationNS field of "default".
- preexisting OSS intentions with now-invalid namespace fields will
delete those intentions on initial election or for wildcard namespaces
an attempt will be made to downgrade them to "default" unless one
exists.
- also allow the '-namespace' CLI arg on all of the intention subcommands
- update lots of docs
We needed to pass a cancellable context into the limiter.Wait instead of context.Background. So I made the func take a context instead of a chan as most places were just passing through a Done chan from a context anyways.
Fix go routine leak in the gateway locator
There are a couple of things in here.
First, just like auto encrypt, any Cluster.AutoConfig RPC will implicitly use the less secure RPC mechanism.
This drastically modifies how the Consul Agent starts up and moves most of the responsibilities (other than signal handling) from the cli command and into the Agent.
Right now this is only hooked into the insecure RPC server and requires JWT authorization. If no JWT authorizer is setup in the configuration then we inject a disabled “authorizer” to always report that JWT authorization is disabled.
While upgrading servers to a new version, I saw that metadata of
existing servers are not upgraded, so the version and raft meta
is not up to date in catalog.
The only way to do it was to:
* update Consul server
* make it leave the cluster, then metadata is accurate
That's because the optimization to avoid updating catalog does
not take into account metadata, so no update on catalog is performed.
And fix the 'value not used' issues.
Many of these are not bugs, but a few are tests not checking errors, and
one appears to be a missed error in non-test code.
A Node Identity is very similar to a service identity. Its main targeted use is to allow creating tokens for use by Consul agents that will grant the necessary permissions for all the typical agent operations (node registration, coordinate updates, anti-entropy).
Half of this commit is for golden file based tests of the acl token and role cli output. Another big updates was to refactor many of the tests in agent/consul/acl_endpoint_test.go to use the same style of tests and the same helpers. Besides being less boiler plate in the tests it also uses a common way of starting a test server with ACLs that should operate without any warnings regarding deprecated non-uuid master tokens etc.
* Fixes#5606: Tokens converted from legacy ACLs get their Hash computed
This allows new style token replication to work for legacy tokens as well when they change.
* tests: fix timestamp comparison
Co-authored-by: Matt Keeler <mjkeeler7@gmail.com>
Found using staticcheck.
binary.Write does not accept int types without a size. The error from binary.Write was ignored, so we never saw this error. Casting the data to uint64 produces a correct hash.
Also deprecate the Default{Addr,Port} fields, and prevent them from being encoded. These fields will always be empty and are not used.
Removing these would break backwards compatibility, so they are left in place for now.
Co-authored-by: Hans Hasselberg <me@hans.io>
The ACL.GetPolicy RPC endpoint was supposed to return the “parent” policy and not always the default policy. In the case of legacy management tokens the parent policy was supposed to be “manage”. The result of us not sending this properly was that operations that required specifically a management token such as saving a snapshot would not work in secondary DCs until they were upgraded.
* testing: replace most goe/verify.Values with require.Equal
One difference between these two comparisons is that go/verify considers
nil slices/maps to be equal to empty slices/maps, where as testify/require
does not, and does not appear to provide any way to enable that behaviour.
Because of this difference some expected values were changed from empty
slices to nil slices, and some calls to verify.Values were left.
* Remove github.com/pascaldekloe/goe/verify
Reduce the number of assertion packages we use from 2 to 1
In the past TLS usage was enforced with these variables, but these days
this decision is made by TLSConfigurator and there is no reason to keep
using the variables.
The version field has been used to decide which multiplexing to use. It
was introduced in 2457293dceec95ecd12ef4f01442e13710ea131a. But this is
6y ago and there is no need for this differentiation anymore.
Three of the checks are temporarily disabled to limit the size of the
diff, and allow us to enable all the other checks in CI.
In a follow up we can fix the issues reported by the other checks one
at a time, and enable them.
The main fix here is to always union the `primary-gateways` list with
the list of mesh gateways in the primary returned from the replicated
federation states list. This will allow any replicated (incorrect) state
to be supplemented with user-configured (correct) state in the config
file. Eventually the game of random selection whack-a-mole will pick a
winning entry and re-replicate the latest federation states from the
primary. If the user-configured state is actually the incorrect one,
then the same eventual correct selection process will work in that case,
too.
The secondary fix is actually to finish making wanfed-via-mgws actually
work as originally designed. Once a secondary datacenter has replicated
federation states for the primary AND managed to stand up its own local
mesh gateways then all of the RPCs from a secondary to the primary
SHOULD go through two sets of mesh gateways to arrive in the consul
servers in the primary (one hop for the secondary datacenter's mesh
gateway, and one hop through the primary datacenter's mesh gateway).
This was neglected in the initial implementation. While everything
works, ideally we should treat communications that go around the mesh
gateways as just provided for bootstrapping purposes.
Now we heuristically use the success/failure history of the federation
state replicator goroutine loop to determine if our current mesh gateway
route is working as intended. If it is, we try using the local gateways,
and if those don't work we fall back on trying the primary via the union
of the replicated state and the go-discover configuration flags.
This can be improved slightly in the future by possibly initializing the
gateway choice to local on startup if we already have replicated state.
This PR does not address that improvement.
Fixes#7339
Errors are values. We can use the error value to identify the 'comparison failed' case which makes the function easier to use and should make it harder to miss handle the error case
Based on work done in https://github.com/hashicorp/memberlist/pull/196
this allows to restrict the IP ranges that can join a given Serf cluster
and be a member of the cluster.
Restrictions on IPs can be done separatly using 2 new differents flags
and config options to restrict IPs for LAN and WAN Serf.
Handling errors at the end of a log switch/case block is somewhat
brittle. This block included a couple cases where errors were ignored,
but it was not obvious the way it was written.
This change moves all error handling into each case block. There is
still potentially one case where err is ignored, which will be handled
in a follow up.
Some of these problems are minor (unused vars), but others are real bugs (ignored errors).
Co-authored-by: Matt Keeler <mkeeler@users.noreply.github.com>
This commit converts the previous error into just a Warn-level log
message. By returning an error when the requested service was not a
gateway, we did not appropriately update envoy because the cache Fetch
returned an error and thus did not propagate the update through proxycfg
and xds packages.
Previously this happened to be using the method on the Server/Client that was meant to allow the ACLResolver to locally resolve tokens. On Servers that had tokens (primary or secondary dc + token replication) this function would lookup the token from raft and return the ACLIdentity. On clients this was always a noop. We inadvertently used this function instead of creating a new one when we added logging accessor ids for permission denied RPC requests.
With this commit, a new method is used for resolving the identity properly via the ACLResolver which may still resolve locally in the case of being on a server with tokens but also supports remote token resolution.
* Return early from updateGatewayServices if nothing to update
Previously, we returned an empty slice of gatewayServices, which caused
us to accidentally delete everything in the memdb table
* PR comment and better formatting
We require any non-wildcard services to match the protocol defined in
the listener on write, so that we can maintain a consistent experience
through ingress gateways. This also helps guard against accidental
misconfiguration by a user.
- Update tests that require an updated protocol for ingress gateways
This now requires some type of protocol setting in ingress gateway tests
to ensure the services are not filtered out.
- small refactor to add a max(x, y) function
- Use internal configEntryTxn function and add MaxUint64 to lib
- Validate that this cannot be set on a 'tcp' listener nor on a wildcard
service.
- Add Hosts field to api and test in consul config write CLI
- xds: Configure envoy with user-provided hosts from ingress gateways
This commit adds the necessary changes to allow an ingress gateway to
route traffic from a single defined port to multiple different upstream
services in the Consul mesh.
To do this, we now require all HTTP requests coming into the ingress
gateway to specify a Host header that matches "<service-name>.*" in
order to correctly route traffic to the correct service.
- Differentiate multiple listener's route names by port
- Adds a case in xds for allowing default discovery chains to create a
route configuration when on an ingress gateway. This allows default
services to easily use host header routing
- ingress-gateways have a single route config for each listener
that utilizes domain matching to route to different services.
This is a collection of refactors that make upcoming PRs easier to digest.
The main change is the introduction of the authmethod.Identity struct.
In the one and only current auth method (type=kubernetes) all of the
trusted identity attributes are both selectable and projectable, so they
were just passed around as a map[string]string.
When namespaces were added, this was slightly changed so that the
enterprise metadata can also come back from the login operation, so
login now returned two fields.
Now with some upcoming auth methods it won't be true that all identity
attributes will be both selectable and projectable, so rather than
update the login function to return 3 pieces of data it seemed worth it
to wrap those fields up and give them a proper name.
Also ensure that WatchSets in tests are reset between calls to watchFired.
Any time a watch fires, subsequent calls to watchFired on the same WatchSet
will also return true even if there were no changes.
Previously, if a blocking query called CheckConnectServiceNodes
before the gateway-services memdb table had any entries,
a nil watchCh would be returned when calling serviceTerminatingGatewayNodes.
This means that the blocking query would not fire if a gateway config entry
was added after the watch started.
In cases where the blocking query started on proxy registration,
the proxy could potentially never become aware of an upstream endpoint
if that upstream was going to be represented by a gateway.
On every service registration, we check to see if a service should be
assassociated to a wildcard gateway-service. This fixes an issue where
we did not correctly check to see if the service being registered was a
"typical" service or not.
* Implements a simple, tcp ingress gateway workflow
This adds a new type of gateway for allowing Ingress traffic into Connect from external services.
Co-authored-by: Chris Piraino <cpiraino@hashicorp.com>
Also reduce the log level of some version checking messages on the server as they can be pretty noisy during upgrades and really are more for debugging purposes.
The test had two racy bugs related to memdb references.
The first was when we initially populated data and retained the FederationState objects in a slice. Due to how the `inmemCodec` works these were actually the identical objects passed into memdb.
The second was that the `checkSame` assertion function was reading from memdb and setting the RaftIndexes to zeros to aid in equality checks. This was mutating the contents of memdb which is a no-no.
With this fix, the command:
```
i=0; while /usr/local/bin/go test -count=1 -timeout 30s github.com/hashicorp/consul/agent/consul -run '^(TestReplication_FederationStates)$'; do i=$((i + 1)); printf "$i "; done
```
That used to break on my machine in less than 20 runs is now running 150+ times without any issue.
Might also fix#7575
* Enable filtering language support for the v1/connect/intentions listing API
* Update website for filtering of Intentions
* Update website/source/api/connect/intentions.html.md
This config entry will be used to configure terminating gateways.
It accepts the name of the gateway and a list of services the gateway will represent.
For each service users will be able to specify: its name, namespace, and additional options for TLS origination.
Co-authored-by: Kyle Havlovitz <kylehav@gmail.com>
Co-authored-by: Chris Piraino <cpiraino@hashicorp.com>
* Add Ingress gateway config entry and other relevant structs
* Add api package tests for ingress gateways
* Embed EnterpriseMeta into ingress service struct
* Add namespace fields to api module and test consul config write decoding
* Don't require a port for ingress gateways
* Add snakeJSON and camelJSON cases in command test
* Run Normalize on service's ent metadata
Sadly cannot think of a way to test this in OSS.
* Every protocol requires at least 1 service
* Validate ingress protocols
* Update agent/structs/config_entry_gateways.go
Co-authored-by: Chris Piraino <cpiraino@hashicorp.com>
Co-authored-by: Freddy <freddygv@users.noreply.github.com>
These changes are necessary to ensure advertisement happens correctly even when datacenters are connected via network areas in Consul enterprise.
This also changes how we check if ACLs can be upgraded within the local datacenter. Previously we would iterate through all LAN members. Now we just use the ServerLookup type to iterate through all known servers in the DC.
This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch:
There are several distinct chunks of code that are affected:
* new flags and config options for the server
* retry join WAN is slightly different
* retry join code is shared to discover primary mesh gateways from secondary datacenters
* because retry join logic runs in the *agent* and the results of that
operation for primary mesh gateways are needed in the *server* there are
some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur
at multiple layers of abstraction just to pass the data down to the right
layer.
* new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers
* the function signature for RPC dialing picked up a new required field (the
node name of the destination)
* several new RPCs for manipulating a FederationState object:
`FederationState:{Apply,Get,List,ListMeshGateways}`
* 3 read-only internal APIs for debugging use to invoke those RPCs from curl
* raft and fsm changes to persist these FederationStates
* replication for FederationStates as they are canonically stored in the
Primary and replicated to the Secondaries.
* a special derivative of anti-entropy that runs in secondaries to snapshot
their local mesh gateway `CheckServiceNodes` and sync them into their upstream
FederationState in the primary (this works in conjunction with the
replication to distribute addresses for all mesh gateways in all DCs to all
other DCs)
* a "gateway locator" convenience object to make use of this data to choose
the addresses of gateways to use for any given RPC or gossip operation to a
remote DC. This gets data from the "retry join" logic in the agent and also
directly calls into the FSM.
* RPC (`:8300`) on the server sniffs the first byte of a new connection to
determine if it's actually doing native TLS. If so it checks the ALPN header
for protocol determination (just like how the existing system uses the
type-byte marker).
* 2 new kinds of protocols are exclusively decoded via this native TLS
mechanism: one for ferrying "packet" operations (udp-like) from the gossip
layer and one for "stream" operations (tcp-like). The packet operations
re-use sockets (using length-prefixing) to cut down on TLS re-negotiation
overhead.
* the server instances specially wrap the `memberlist.NetTransport` when running
with gateway federation enabled (in a `wanfed.Transport`). The general gist is
that if it tries to dial a node in the SAME datacenter (deduced by looking
at the suffix of the node name) there is no change. If dialing a DIFFERENT
datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh
gateways to eventually end up in a server's :8300 port.
* a new flag when launching a mesh gateway via `consul connect envoy` to
indicate that the servers are to be exposed. This sets a special service
meta when registering the gateway into the catalog.
* `proxycfg/xds` notice this metadata blob to activate additional watches for
the FederationState objects as well as the location of all of the consul
servers in that datacenter.
* `xds:` if the extra metadata is in place additional clusters are defined in a
DC to bulk sink all traffic to another DC's gateways. For the current
datacenter we listen on a wildcard name (`server.<dc>.consul`) that load
balances all servers as well as one mini-cluster per node
(`<node>.server.<dc>.consul`)
* the `consul tls cert create` command got a new flag (`-node`) to help create
an additional SAN in certs that can be used with this flavor of federation.
Daniel Nephin