Add a new config option for Vault Agent's JWT auto auth
`remove_jwt_after_reading`, which defaults to true. Can stop
Agent from attempting to delete the file, which is useful in k8s
where the service account JWT is mounted as a read-only file
and so any attempt to delete it generates spammy error logs.
When leaving the JWT file in place, the read period for new
tokens is 1 minute instead of 500ms to reflect the assumption
that there will always be a file there, so finding a file does not
provide any signal that it needs to be re-read. Kubernetes
has a minimum TTL of 10 minutes for tokens, so a period of
1 minute gives Agent plenty of time to detect new tokens,
without leaving it too unresponsive. We may want to add a
config option to override these default periods in the future.
Co-authored-by: Tom Proctor <tomhjp@users.noreply.github.com>
* agent/azure: adds ability to use specific user assigned managed identity for auto auth
* add changelog
* change wording in error and docs
* Update website/content/docs/agent/autoauth/methods/azure.mdx
Co-authored-by: Theron Voran <tvoran@users.noreply.github.com>
* Update website/content/docs/agent/autoauth/methods/azure.mdx
Co-authored-by: Tom Proctor <tomhjp@users.noreply.github.com>
* docs formatting
Co-authored-by: Theron Voran <tvoran@users.noreply.github.com>
Co-authored-by: Tom Proctor <tomhjp@users.noreply.github.com>
This patch adds a new /agent/v1/metrics that will return metrics on the
running Vault agent. Configuration is done using the same telemetry
stanza as the Vault server. For now default runtime metrics are
returned with a few additional ones specific to the agent:
- `vault.agent.auth.failure` and `vault.agent.auth.success` to monitor
the correct behavior of the auto auth mechanism
- `vault.agent.proxy.success`, `vault.agent.proxy.client_error` and
`vault.agent.proxy.error` to check the connection with the Vault server
- `vault.agent.cache.hit` and `vault.agent.cache.miss` to monitor the
cache
Closes https://github.com/hashicorp/vault/issues/8649
Co-authored-by: Theron Voran <tvoran@users.noreply.github.com>
Unlike the other libraries that were migrated, there are no usages of
this lib in any of our plugins, and the only other known usage was in
go-kms-wrapping, which has been updated. Aliasing it like the other libs
would still keep the aws-sdk-go dep in the sdk module because of the
function signatures. So I've simply removed it entirely here.
* Update Agent Auth with GCP to use new SignJWT endpoint
* use iamcredentials name instead of renaming the package on import
* add changelog
* Update changelog/11473.txt
Co-authored-by: Theron Voran <tvoran@users.noreply.github.com>
Co-authored-by: Theron Voran <tvoran@users.noreply.github.com>
The existing code would retain the previous backoff value even after the
system had recovered. This PR fixes that issue and improves the
structure of the backoff code.
* agent: return a non-zero exit code on error
* agent/template: always return on template server error, add case for error_on_missing_key
* agent: fix tests by updating Run params to use an errCh
* agent/template: add permission denied test case, clean up test var
* agent: use unbuffered errCh, emit fatal errors directly to the UI output
* agent: use oklog's run.Group to schedule subsystem runners (#9761)
* agent: use oklog's run.Group to schedule subsystem runners
* agent: clean up unused DoneCh, clean up agent's main Run func
* agent/template: use ts.stopped.CAS to atomically swap value
* fix tests
* fix tests
* agent/template: add timeout on TestRunServer
* agent: output error via logs and return a generic error on non-zero exit
* fix TestAgent_ExitAfterAuth
* agent/template: do not restart ct runner on new incoming token if exit_after_auth is set to true
* agent: drain ah.OutputCh after sink exits to avoid blocking on the channel
* use context.WithTimeout, expand comments around ordering of defer cancel()
Adds debug and warn logging around AWS credential chain generation,
specifically to help users debugging auto-unseal problems on AWS, by
logging which role is being used in the case of a webidentity token.
Adds a deferred call to flush the log output as well, to ensure logs
are output in the event of an initialization failure.
* Sync up Agent and API's renewers.
This introduces a new type, LifetimeWatcher, which can handle both
renewable and non-renewable secrets, modeled after the version in Agent.
It allows the user to select behavior, with the new style being the
default when calling Start(), and old style if using the legacy Renew()
call.
No tests have been modified (except for reflect issues) and no other
code has been modified to make sure the changes are backwards
compatible.
Once this is accepted I'll pull the Agent version out.
* Move compat flags to NewRenewer
* Port agent to shared lifetime watcher lib
* Vault Agent Template: parse templates (#7540)
* add template config parsing, but it's wrong b/c it's not using mapstructure
* parsing consul templates in agent config
* add additional test to configuration parsing, to cover basics
* another test fixture, rework simple test into table
* refactor into table test
* rename test
* remove flattenKeys and add other test fixture
* Update command/agent/config/config.go
Co-Authored-By: Jim Kalafut <jkalafut@hashicorp.com>
* return the decode error instead of swallowing it
* Update command/agent/config/config_test.go
Co-Authored-By: Jim Kalafut <jkalafut@hashicorp.com>
* go mod tidy
* change error checking style
* Add agent template doc
* TemplateServer: render secrets with Consul Template (#7621)
* add template config parsing, but it's wrong b/c it's not using mapstructure
* parsing consul templates in agent config
* add additional test to configuration parsing, to cover basics
* another test fixture, rework simple test into table
* refactor into table test
* rename test
* remove flattenKeys and add other test fixture
* add template package
* WIP: add runner
* fix panic, actually copy templates, etc
* rework how the config.Vault is created and enable reading from the environment
* this was supposed to be a part of the prior commit
* move/add methods to testhelpers for converting some values to pointers
* use new methods in testhelpers
* add an unblock channel to block agent until a template has been rendered
* add note
* unblock if there are no templates
* cleanups
* go mod tidy
* remove dead code
* simple test to starT
* add simple, empty templates test
* Update package doc, error logs, and add missing close() on channel
* update code comment to be clear what I'm referring to
* have template.NewServer return a (<- chan) type, even though it's a normal chan, as a better practice to enforce reading only
* Update command/agent.go
Co-Authored-By: Jim Kalafut <jkalafut@hashicorp.com>
* update with test
* Add README and doc.go to the command/agent directory (#7503)
* Add README and doc.go to the command/agent directory
* Add link to website
* address feedback for agent.go
* updated with feedback from Calvin
* Rework template.Server to export the unblock channel, and remove it from the NewServer function
* apply feedback from Nick
* fix/restructure rendering test
* Add pointerutil package for converting types to their pointers
* Remove pointer helper methods; use sdk/helper/pointerutil instead
* update newRunnerConfig to use pointerutil and empty strings
* only wait for unblock if template server is initialized
* drain the token channel in this test
* conditionally send on channel
* adding auto auth for cert auth methods
* Adding Docs for Cert Auto-auth method
* fixes errors in docs
* wrong documentation title
* repariting a few typos in the docs and being very clear about where the certificates should be configured
* clarifying the name parameter documentation
* Fixes gofmt issues in command/agent.go
* Fix typo in docs
* Add nil check to Config for cert auto-auth and add test with and without a specified name
The result will still pass gofmtcheck and won't trigger additional
changes if someone isn't using goimports, but it will avoid the
piecemeal imports changes we've been seeing.
