* Avastancu/joannaakl/service container error log (#2110)
* adding support for a service container docker logs
* Adding Unit test to ContainerOperationProvider
* Adding another test to ContainerOperationProvider
* placed the docker logs output in dedicated ##group section
* Removed the exception thrown if the service container was not healthy
* Removed duplicated logging to the executionContext
* Updated the container logs sub-section message
* Print service containers only if they were healthy
Unhealthy service logs are printed in ContainerHealthCheckLogs called prior to this step.
* Removed recently added method to inspect docker logs
The method was doing the same thing as the existing DockerLogs method.
* Added execution context error
This will make a failed health check more visible in the UI without disrupting the execution of the program.
* Removing the section 'Waiting for all services to be ready'
Since nested subsections are not being displayed properly and we already need one subsection per service error.
* Update src/Runner.Worker/Container/DockerCommandManager.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/TestHostContext.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Change the logic for printing Service Containers logs
Service container logs will be printed in the 'Start containers' section only if there is an error.
Healthy services will have their logs printed in the 'Stop Containers' section.
* Removed unused import
* Added back section group.
* Moved service containers error logs to separate group sections
* Removed the test testing the old logic flow.
* Remove unnecessary 'IsAnyUnhealthy' flag
* Remove printHello() function
* Add newline to TestHostContext
* Remove unnecessary field 'UnhealthyContainers'
* Rename boolean flag indicating service container failure
* Refactor healthcheck logic to separate method to enable unit testing.
* Remove the default value for bool variable
* Update src/Runner.Worker/ContainerOperationProvider.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Runner.Worker/ContainerOperationProvider.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Rename Healthcheck back to ContainerHealthcheck
* Make test sequential
* Unextract the container error logs method
* remove test asserting thrown exception
* Add configure await
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Add back test asserting exception
* Check service exit code if there is no healtcheck configured
* Remove unnecessary healthcheck for healthy service container
* Revert "Check service exit code if there is no healtcheck configured"
This reverts commit fec24e8341.
Co-authored-by: Ava S <avastancu@github.com>
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Do not fail service containers without the healthcheck
Co-authored-by: JoannaaKL <joannaakl@github.com>
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* adding support for a service container docker logs
* Adding Unit test to ContainerOperationProvider
* Adding another test to ContainerOperationProvider
* placed the docker logs output in dedicated ##group section
* Removed the exception thrown if the service container was not healthy
* Removed duplicated logging to the executionContext
* Updated the container logs sub-section message
* Print service containers only if they were healthy
Unhealthy service logs are printed in ContainerHealthCheckLogs called prior to this step.
* Removed recently added method to inspect docker logs
The method was doing the same thing as the existing DockerLogs method.
* Added execution context error
This will make a failed health check more visible in the UI without disrupting the execution of the program.
* Removing the section 'Waiting for all services to be ready'
Since nested subsections are not being displayed properly and we already need one subsection per service error.
* Update src/Runner.Worker/Container/DockerCommandManager.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/TestHostContext.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Change the logic for printing Service Containers logs
Service container logs will be printed in the 'Start containers' section only if there is an error.
Healthy services will have their logs printed in the 'Stop Containers' section.
* Removed unused import
* Added back section group.
* Moved service containers error logs to separate group sections
* Removed the test testing the old logic flow.
* Remove unnecessary 'IsAnyUnhealthy' flag
* Remove printHello() function
* Add newline to TestHostContext
* Remove unnecessary field 'UnhealthyContainers'
* Rename boolean flag indicating service container failure
* Refactor healthcheck logic to separate method to enable unit testing.
* Remove the default value for bool variable
* Update src/Runner.Worker/ContainerOperationProvider.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Runner.Worker/ContainerOperationProvider.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Rename Healthcheck back to ContainerHealthcheck
* Make test sequential
* Unextract the container error logs method
* remove test asserting thrown exception
* Add configure await
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Test/L0/Worker/ContainerOperationProviderL0.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Add back test asserting exception
* Check service exit code if there is no healtcheck configured
* Remove unnecessary healthcheck for healthy service container
* Revert "Check service exit code if there is no healtcheck configured"
This reverts commit fec24e8341.
Co-authored-by: Ava S <avastancu@github.com>
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Adding a new vars context for non-secret variables
* Fix test case
* Trigger checks
* Remove variables from env context and environment varibale
* remove extra references
* Add prefix handling to configuration variables
* Fix test cases
* Consume variables using vars in context data
* removed action_yaml changes
* Added ability to run Dockerfile.SUFFIX ContainerAction
* Extracted IsDockerFile method
* reformatted, moved from index to Last()
* extracted IsDockerfile to DockerUtil with L0
* added check for IsDockerfile to account for docker://
* updated test to clearly show path/dockerfile:tag
* fail if Data.Image is not Dockerfile or docker://[image]
* Setup noops for JobPrepare and JobCleanup hooks
* Add container jobstarted and jobcomplete hooks
* Run 'index.js' instead of specific command hooks
* Call jobprepare with command arg
* Use right command name (hardcoded)
Co-authored-by: Nikola Jokic <nikola-jokic@users.noreply.github.com>
* Invoke hooks with arguments
* Add PrepareJob hook to work with jobcontainers
Co-authored-by: Nikola Jokic <nikola-jokic@users.noreply.github.com>
* Rename methods
* Use new hookcontainer to run prep and clean hooks
* Get path from ENV
* Use enums
* Use IOUtils.cs
* Move container files to folder
* Move namespaces
* Store "state" between hooks
* Remove stdin stream in containerstephosts
* Update Constants.cs
* Throw if stdin fails
* Cleanup obvious nullrefs and unused vars
* Cleanup containerhook directory
* Call step exec hook
* Fix windows build
* Remove hook from hookContainer
* Rename file
* More renamings
* Add TODOs
* Fix env name
* Fix missing imports
* Fix imports
* Run script step on jobcontainer
* Enable feature if env is set
* Update ContainerHookManager.cs
* Update ContainerHookManager.cs
* Hooks allowed to work even when context isn't returned
* Custom hooks enabled flag and additional null checks
* New line at the end of the FeatureFlagManager.cs
* Code refactoring
* Supported just in time container building or pulling
* Try mock-build for osx
* Build all platforms
* Run mock on self-hosted
* Remove GITHUB prefix
* Use ContainerHooksPath instead of CustomHooksPath
* Null checks simplified
* Code refactoring
* Changing condition for image builing/pulling
* Code refactoring
* TODO comment removed
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Call container step if FF is on
* Rename run script function
* Use JToken instead of dynamic
* Add TODO
* Small refactoring + renames + TODOs
* Throw on DetermineNodeRuntimeVersion
* Fix formatting
* Add run-container-step
* Supported nodeJS in Alpine containers
* Renamed Alpine to IsAlpine in HookResponse
* Method for checking platform for alpine container
* Added container hooks feature flag check
* Update IsHookFeatureEnabled with new params
* Rename featureflag method
* Finish rename
* Set collection null values to empty arrays when JSON serialising them
* Disable FF until we merge
* Update src/Runner.Worker/Container/ContainerHooks/HookContainer.cs
* Fix method name
* Change hookargs to superclass from interface
* Using only Path.Combine in GenerateResponsePath
* fix merge error
* EntryPointArgs changed to list of args instead of one args string
* Changed List to IEnumerable for EntryPointArgs and MountVolumes
* Get ContainerRuntimePath for JobContainers from hooks
* Read ContainerEnv from response file
* Port mappings saved after creating services
* Support case when responseFile doesn't exist
* Check if response file exists
* Logging in ExecuteHookScript
* Save hook state after all 4 hooks
* Code refactoring
* Remove TODO
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Remove second TODO
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Removing container env changes
* Removing containerEnv and dockerManager
* Delete mock-build.yml
* Update IOUtil.cs
* Add comment about containerhooks
* Fix merge mistake
* Remove solved todo
* Determine which shell to use for hooks scenario
* Overload for method ExecuteHookScript with prependPath as arg
* Adding HostContext to the GetDefaultShellForScript call
* prependPath as a mandatory parameter
* Improve logging for hooks
* Small changes in logging
* Allow null for ContainerEntryPointArgs
* Changed log messages
* Skip setting EntryPoint and EntryPointArgs if hooks are enabled
* Throw if IsAlpine is null in PrepareJob
* Code refactoring - added GetAndValidateResponse method
* Code refactoring
* Changes in exception message
* Only save hookState if returned
* Use FF from server
* Empty line
* Code refactoring
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Send null instead of string empty
* Remove TODO
* Code refactoring and some small changes
* Allow Globals to be null to pass L0
* Fix setup in StepHostL0
* Throw exception earlier if response file doesn't exist and prepare_job hook is running
* Refactoring GetResponse method
* Changing exception message if response file is not found
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Chaning exception message if isAlpine is null for prepare_job hook
* Rename hook folder
* Fail if compatible hookfile not found
* Use .Value instead of casting bool? to bool
* Format spacing
* Formatting
* User user and system mvs
* Use variables instead of entire context in featuremanager
* Update stepTelemetry if step uses containerhooks
* Restore import
* Remove unneccessary field from HookContainer
* Refactor response context and portmappings
* Force allow hooks if FF is on
* Code refactoring
* Revert deleting usings
* Better hookContainer defaults and use correct portmapping list
* Make GetDefaultShellForScript a HostContext extension method
* Generic hookresponse
* Code refactoring, unnecessary properties removed - HookContainer moved to the HookInput.cs
* Remove empty line
* Code refactoring and better exception handling
* code refactor, removing unnecessary props
* Move hookstate to global ContainerHookState
* Trace exception before we throw it for not losing information
* Fix for null ref exception in GetResponse
* Adding additional check for null response in prepareJob hook
* Refactoring GetResponse with additional check
* Update error messages
* Ports in ResponseContainer changed from IList to IDictionary
* Fix port format
* Include dockerfile
* Send null Registry obj if there's nothing in it
* Minor formatting
* Check if hookIndexPath exists relocated to the ContainerHookManager
* Code refactoring - ValidateHookExecutable added to the ContainerHookManager
* check if ContainerHooksPath when AllowRunnerContainerHooks is on
* Submit JSON telemetry instead of boolean
* Prefix step hooks with "run"
* Rename FeatureManager
* Fix flipped condition
* Unify js shell path getter with ps1 and sh getter
* Validate on run, not on instantiation of manager
* Cleanup ExecuteAsync methods
* Handle exception in executeHookScript
* Better exception types
* Remove comment
* Simplify boolean
* Allow jobs without jobContainer to run
* Use JObject instead of JToken
* Use correct Response type
* Format class to move cleanupJobAsync to the end of public methods
* Rename HookIndexPath to HookScriptPath
* Refactor methods into expression bodies
* Fix args class hierarchy
* Fix argument order
* Formatting
* Fix nullref and don't swallow stacktrace
* Whilelist HookArgs
* Use FF in FeatureManager
* Update src/Runner.Worker/ContainerOperationProvider.cs
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
* Update src/Runner.Worker/ActionRunner.cs
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* Update src/Runner.Worker/ActionRunner.cs
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* Only mount well known dirs to job containers
* Get trace from hostcontext
* Use hook execution for setting telemetry
Co-authored-by: Nikola Jokic <nikola.jokic@akvelon.com>
Co-authored-by: Nikola Jokic <nikola-jokic@users.noreply.github.com>
Co-authored-by: Nikola Jokic <97525037+nikola-jokic@users.noreply.github.com>
Co-authored-by: Stefan Ruvceski <stefan.ruvceski@akvelon.com>
Co-authored-by: ruvceskistefan <96768603+ruvceskistefan@users.noreply.github.com>
Co-authored-by: Thomas Boop <thboop@github.com>
Co-authored-by: stefanruvceski <ruvceskistefan@github.com>
Co-authored-by: Tingluo Huang <tingluohuang@github.com>
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* composite action github.action_path set based on the StepHost
* in progress on updating github context for input template
* Fixed updating the context data for evaluation
* refactored logic so it is a little cleaner
* removed resolving the action_path in CompositeActionHandler
* removed added DeepClone
* added feature flag and modified the dict in place
* refactored step host to change context data. Added L0
* repaired spaces
* moved logic from step host to execution context, added recursive translation
* removed empty lines
* moved to extension methods
* Update src/Test/L0/Worker/StepHostL0.cs
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Added continue on error to composite action
* changed from boolean-strategy-context -> boolean-steps-context for action_yaml
* refactored composite handler to always set outcome
* retrigger checks
* fixed typo in ??= operator
* boolean-steps-context accepts the same context as string-steps-context
* setting the outcome only on continue-on-error
* moved continue on error logic to the execution context
* Added L0 table tests for continue-on-error ExecutionContext
* Added missing mocks on StepsRunnerL0 for this update
* removed empty line and added one line separating the call
* Removed empty line
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
Co-authored-by: Ferenc Hammerl <31069338+fhammerl@users.noreply.github.com>
* Set GH actions Node version to 16 if FF is on unless user opted out
* Add L0s (WIP)
* Wrap tests into theory
* Only check for node12 actions
* Refactor node version picking
* Issue 1596: runner throws nullref exception when writting env var
* Adding tests for missing new line after EOF marker
* Changing newline to new line
* First prototype of step summary environment variable
* Fix file contention issue
* Try to simplify cleaning up file references
* use step id as md file name, queue file attachment
* separate logic into attachment summary func
* Fix indentation
* Add (experimental) feature flag support
* reorganize summary upload determination logic
* file i/o exception handling + pr feedback
* Revert changes for now to reintroduce them later
* Add skeleton SetStepSummaryCommand
* Update step summary feature flag name
* Port ShouldUploadAttachment from previous iteration
* Port QueueStepSummaryUpload from previous iteration
* Improve exception handling when uploading attachment
* Add some minor logging improvements
* Refuse to upload files larger than 128k
* Implement secrets scrubbing
* Add TODO comment to remove debugging temp files
* Add first tests
* Add test for secret masking
* Add some naming/style fixes suggested in feedback
* inline check for feature flag
* Inline method for style consistency
* Make sure that scrubbed file doesn't exist before creating it
* Rename SetStepSummaryCommand to CreateStepSummaryCommand
* Fix error handling messages
* Fix file command name when registering extension
* Remove unnecessary file deletion
Co-authored-by: Rob Herley <robherley@github.com>
* conditional support for composite actions
* Fix Conditional function evaluation
* Push launch.json temporarily
* Revert "Push launch.json temporarily"
* rename context
* Cleanup comments
* fix success/failure functions to run based on pre/main steps
* idea of step_status
* change to use steps context, WIP
* add inputs to possible if condition expressions
* use action_status
* pr cleanup
* Added right stages
* Test on stage in conditional functions
* Fix naming and formatting
* Fix tests
* Add success and failure L0s
* Remove comment
* Remove whitespace
* Undo formatting
* Add L0 for step-if parsing
* Add ADR
Co-authored-by: Thomas Boop <thboop@github.com>
* Prevent stopTokens that are workflow commands
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* Check context for env var too
* Accept true, 1 and $true instead of just "true"
* Setup ExpressionValues in tests
* Update src/Runner.Common/Constants.cs
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* Separate success and fail tests for invalid token
* Fix envcontext for tests
Co-authored-by: Thomas Boop <52323235+thboop@users.noreply.github.com>
* Add Notice Command
* Add Feature Flag For Enhanced Annotations
* Fix tests
* Add validation for columns and lines
* Fix order to match service
* Remove console.write
* Make Validation Better
* Cleanup
* Handle empty/whitespace strings
* Add more validation for line/column ranges
* Make Validation Debug, Not Throw
* Change casing to 🐫 from 🐍
* Give notice a well known tag
* Cleanup
* Sanitize invalid commands rather than fail
This PR changes GITHUB_ACTION to use the step ContextName, instead of refname. The behavior is behind a feature flag. Refname is an otherwise deprecated property.
Primary motivation: For composite actions, we need a distinct GITHUB_ACTION for each nested step. This PR adds code to generate a default context name for nested steps.
For nested steps, GITHUB_ACTION will be set to "{ScopeName}.{ContextName}" to ensure no collisions.
A corresponding change will be made on the server so context name is never empty. Generated context names will start with "__".
A follow-up PR is required to avoid tracking "step" context values (outputs/conclusion/result) for generated context names. Waiting on telemetry from the server to confirm it's safe to assume leading "__" is a generate context name.
* Composite Action Run Steps
* Env Flow => Able to get env variables and overwrite current env variables => but it doesn't 'stick'
* clean up
* Clean up trace messages + add Trace debug in ActionManager
* Add debugging message
* Optimize runtime of code
* Change String to string
* Add comma to Composite
* Change JobSteps to a List, Change Register Step function name
* Add TODO, remove unn. content
* Remove unnecessary code
* Fix unit tests
* Fix env format
* Remove comment
* Remove TODO message for context
* Add verbose trace logs which are only viewable by devs
* Initial Start for FileTable stuff
* Progress towards passing FileTable or FileID or FileName
* Sort usings in Composite Action Handler
* Change 0 to location
* Update context variables in composite action yaml
* Add helpful error message for null steps
* Pass fileID to all children token of root action token
* Change confusing term context => templateContext, Eliminate _fileTable and only use ExecutionContext.FileTable + update this table when need be
* Remove unnessary FileID attribute from CompositeActionExecutionData
* Clean up file path for error message
* Remove todo
* Fix Workflow Step Env overiding Parent Env
* Remove env in composite action scope
* Clean up
* Revert back
* revert back
* add back envToken
* Remove unnecessary code
* Add file length check
* Clean up
* Figure out how to handle set-env edge cases
* formatting
* fix unit tests
* Fix windows unit test syntax error
* Fix period
* Sanity check for fileTable add + remove unn. code
* revert back
* Add back line break
* Fix null errors
* Address situation if FileTable is null + add sanity check for adding file to fileTable
* add line
* Revert
* Fix unit tests to instantiate a FileTable
* Fix logic for trimming manifestfile path
* Add null check
* Add filetable to testing file, remove ? since we know filetable should never be non null
* Composite Action Run Steps
* Env Flow => Able to get env variables and overwrite current env variables => but it doesn't 'stick'
* clean up
* Clean up trace messages + add Trace debug in ActionManager
* Add debugging message
* Optimize runtime of code
* Change String to string
* Add comma to Composite
* Change JobSteps to a List, Change Register Step function name
* Add TODO, remove unn. content
* Remove unnecessary code
* Fix unit tests
* Fix env format
* Remove comment
* Remove TODO message for context
* Add verbose trace logs which are only viewable by devs
* Sort usings in Composite Action Handler
* Change 0 to location
* Update context variables in composite action yaml
* Add helpful error message for null steps
* Fix Workflow Step Env overiding Parent Env
* Remove env in composite action scope
* Clean up
* Revert back
* revert back
* add back envToken
* Remove unnecessary code
* Figure out how to handle set-env edge cases
* formatting
* fix unit tests
* Fix windows unit test syntax error
