* Set an explicit timeout on ResultsHttpClient
* Hook up retry and standard message handler to ResultsHttpClient
* Remove explicit timeout constructor
* Fix linter
* Only upload to Results with new job message type
* No need to have separate websocketFeedServer
* Linting fix
* Update src/Runner.Common/JobServerQueue.cs
Co-authored-by: Tingluo Huang <[email protected]>
* add connection timeout
* Consolidate initializing webclient to result client
* Retry websocket delivery for console logs
* Linter fix
* Do not give up for good, reconnect again in 10 minutes
* Has to reset delivered
* Only first time retry 3 times to connect to websocket
---------
Co-authored-by: Tingluo Huang <[email protected]>
* Also send Steps update to Results service
* Refactor to separate results server from current job server
* If hit any error while uploading to Results, skip Results upload
* Add proxy authentication and buffer request for WinHttpHandler
* Remove unnecessary null guard
* Also send Results telemetry when step update fails
* IResultsServer is not disposable
* Revert "Revert "Uploading step logs to Results as well (#2422)" (#2437)"
This reverts commit 8c096baf49.
* Properly guard the upload to results feature
* Delete skipped file if deletesource is true
* Rename queue to results queue
* Add results contracts
* Add Results logging handling
* Adding calls to create and finalize append blob
* Modifications for azurite upload
* Only call upload complete on final section and remove size
* Make method specific to step log so we can support job log later
* Change contract for results
* Add totalline count to the result log upload file
* Actually pass lineCount to Results Service
* Fix typos
* Code cleanup
* Fixing typos
* Apply suggestions from code review
Co-authored-by: Konrad Pabjan <[email protected]>
---------
Co-authored-by: Brittany Ellich <[email protected]>
Co-authored-by: Konrad Pabjan <[email protected]>
* Remove unnecessary timelineId and timelineRecordId and use Guid stepId
* Log upload error to kusto
* Remove try-catch
* Using a well known telmetry record to avoid replacing issues
* fix Guid format
During job run we may fail to resolve actions download info, and this
stack is fully controlled by GitHub actions so it should be counted as
infrastructure failure instead of user failure.
