- add `LocalStorageState` to worker status reports
- update worker service handler to process `LocalStorageState`. If is empty for workers older workers, we set the `LocalStorageState` to `unknown`
- update field mask paths to support `LocalStorageState`
At some point we migrated to grpc ConnectParams to control how often
retries were happening. One value that we didn't set was the
MinConnectTimeout. It turns out that if you don't set this, it won't use
the default; instead if will actually use a value of zero, which gRPC
internally ups to something that seems to be around 25ms to meet a
minimum threshold. This is not documented in the function comments.
This commit simply sets the value of MinConnectTimeout to the unexported
gRPC default.
This has been superceded by the nodeenrollment-based KMS worker auth
mechanism. This commit only removes the ability to connect with this
mechanism; some upgrade logic will remain intact for a bit and
can be removed later.
* Minor test fixes
The auth rotation test fix allows us to keep checking if we did not yet
see the expected bytes in case the call to lookup succeeds but we are
late in rotating due to parallelism.
The api credentials test fix allows us to check if at _least_ one
rotation has happened. That way if two rotations happen before we check
again due to parallelism, this is deemed a successful result.
* Add timeout to auth rotation test
The nhooyr/websocket library now has a release with the new default
behavior that removes the message size limit. This allows us to switch
off of the fork and back to the upstream.
In the same release that includes the fix we need for the boundary
client, the wspb helper package was removed from websocket. This package
contained some functions to read and write protobuf messages. However,
it was using the deprecated github.com/golang/protobuf/proto library. To
avoid this dependancy in the library they opted to just remove wspb.
Thus this provides a replacement implementation that uses
google.golang.org/protobuf/proto. Notably the newer proto library seems
to lack an equivalent of proto.NewBuffer so writes to the websocket and
not able to leverage a sync.Pool of byte.Buffers to reduce allocations.
Refs: 88bb253554
See: https://github.com/nhooyr/websocket/commit/ac385120c6e34fa6584f3856d5db949a21bbb65e
* fix: monitorUpstreamConnectionState CPU consumption
`monitorUpstreamConnectionState()` is a goroutine that listens for GRPC client connection changes using GRPC's `WaitForStateChange` method. `monitorUpstreamConnectionState()` was consuming the a lot of CPU. It was continuously running and not actually waiting for state changes. The `state` for `monitorUpstreamConnectionState()` was not getting correctly updated which caused the loop to continuously run instead of waiting for new state changes.
Update GRPC `WaitForStateChange` method to use the same `state` that gets updated.
In preparation for putting proxy code in the api package this creates some necessary fields in the target output and it pre-moves credential parsing and proxy handshake code to their new places. This will make the diff for the actual proxy code move quite a bit smaller.
Co-authored-by: Johan Brandhorst-Satzkorn <johan.brandhorst@gmail.com>
With a suitably slow system -- say, if you're waiting on a Windows
Terminal to output text inside a virtual machine -- the expected order
of things can be thrown off. One specific way this can happen is that
the worker can attempt to dial and have its dial fail because of the
slowness causing the dial context to time out. If, because of the
slowness, we also haven't released the log gate yet, then the event with
the failure information will be queued, along with the context that was
used.
Unfortunately, in some of these error cases, the context that was used
was the dial context instead of the system base context. In many other
places in the function it was the system base context so this is just a
mismatch, probably from code written at different times.
Normally this wouldn't be a problem as we'd fall back to the underlying
logger, but when we release the log gate, things happen differently:
it's a synchronous function and on error it causes us to abandon
system startup entirely.
This commit fixes the issue with using the incorrect context. It's an
open question whether we should change the behavior around errors when
replaying queued events, falling back to the underlying logger instead
of erroring.
* fix (events): convert multierror to stdlib error
multierror errors will be converted to a stdlib error (because
multierror doesn't support json marshaling).
* refact (errors): convert from multierror to stdlib errors.Join(...)
* refact: convert from pkg/errors to stdlib error wrapping
* feat: Worker Connectivity Health Check
Add worker upstream connection state. API calls to the worker's ops `/heath?worker_info=1` endpoint should now return a new filed called `upstream_connection_state`
```
{
"worker_process_info":{
"state":"active",
"active_session_count":0,
"upstream_connection_state":"READY"
}
}
```
The states returned will be GRPC connection states: https://grpc.github.io/grpc/core/md_doc_connectivity-semantics-and-api.html
There could be multiple upstream addresses but GRPC manual resolver determines the state of the connection.
* Adding explicit MPL license for sub-package.
This directory and its subdirectories (packages) contain files licensed with the MPLv2 `LICENSE` file in this directory and are intentionally licensed separately from the BSL `LICENSE` file at the root of this repository.
* Adding explicit MPL license for sub-package
This directory and its subdirectories (packages) contain files licensed with the MPLv2 `LICENSE` file in this directory and are intentionally licensed separately from the BSL `LICENSE` file at the root of this repository.
* Updating the license from MPL to Business Source License
Going forward, this project will be licensed under the Business Source License v1.1. Please see our blog post for more details at https://hashi.co/bsl-blog, FAQ at www.hashicorp.com/licensing-faq, and details of the license at www.hashicorp.com/bsl.
* Update copyright file headers to BUS-1.1
* Rerun make gen
This will pick up the last of the license changes
* Revert "Rerun make gen"
This reverts commit dbe0968bffd92745d8663c6940debae098395a6b.
* Remove copywrite from generation script
The current version of copywrite is not sophisticated enough
to only apply the copywrite changes to certain subdirectories,
so to avoid accidentally relicensing our api and sdk modules,
we will stop running the copywrite script until such time
that it can do the right thing.
* License protobuf API as MPL
The protobuf API files are used to generate the SDK,
since we want the SDK to be MPL, the protobuf files
must also be MPL. This extends to all protobuf files
whose sources end up in the SDK directory, including:
- internal/proto/controller/api
- internal/proto/controller/custom_options
- internal/proto/plugin
It also includes the generated service directory in
internal/gen/controller/api/services, since they are
generated from the protobuf sources used for the SDK.
* Change website license file name
This makes the naming scheme consistent with the rest
of the repository.
* Add note on licensing to README
* Set copywrite license to BUSL
* Revert "Add note on licensing to README"
This reverts commit 4d865a91a7a4f11605f0dc4609355d7e3c3cfffe.
---------
Co-authored-by: hashicorp-copywrite[bot] <110428419+hashicorp-copywrite[bot]@users.noreply.github.com>
Co-authored-by: Johan Brandhorst-Satzkorn <johan.brandhorst@gmail.com>
Most of the changes here are changes or enhancements that are useful for
tests. The crux of the issue was that I had pulled up one var to the
package level so it could be read from tests, but that var was not being
reset between tests. It's not even used in tests anymore, so I put it
back to function-local.
* Add more logging to worker auth rotation and test flags
This makes it much easier to figure out when and why rotation is
happening and control it for tests/debugging.
* Update auth rotation logic
Because we were looking at validity period of client certs, which may
not have matched the actual rotation interval, it was possible (mostly
in test scenarios) for the rotation to happen multiple times on the
roots before we actually hit the worker rotation interval.
This changes how the interval is calculated by looking at the difference
between not-after values on the certs to understand the rotation
interval being used by the controller. We then use that as the starting
point for calculating a next-rotation time.
This makes it much easier to figure out when and why rotation is
happening and control it for tests/debugging.
Remove cert skew
Cherry-picked from release/0.12.x
Elim TsiagbeyandDamian Debkowski
Irena Rindos
Michael LiandElim Tsiagbey
dani
Jeff Mitchell
Timothy Messier
Jim
hashicorp-copywrite[bot]andJohan Brandhorst-Satzkorn
Todd
Louis Ruch