Repository navigation
No retries on REST API time-outs #295
Description
Describe the bug
Recently the ADF REST API seems to time-out frequently, causing the deployment to fail. While API time-outs are a Microsoft issue, I do think a retry on the API calls should be essential. For me this is happening multiple times a day now.
Since ADF is only a rather small part of the deployment, having to re-run the whole thing for multiple projects at a time can be quite daunting.
I do would understand if you'd see this as an enhancement rather than a bug though. Please let me know how you see this issue!
To Reproduce
N/A, intermittent issue with Microsoft
Expected behaviour
When an API time-out occurs, the script should re-attempt the API call a few times, up to a configurable amount. This prevents deployments to fail.
Screenshots
Version of PowerShell module of:
- 0.101.000
Provide all files to reproduce the issue
2023-03-29T08:40:02.1628240Z Start deploying object: [pipeline].[pipeline_name] (2 dependency/ies)
2023-03-29T08:41:42.3155920Z New-AzResource : �[31;1mOperation failed because a request timed out.�[0m
2023-03-29T08:41:42.3158039Z �[31;1mAt /home/AzDevOps/.local/share/powershell/Modules/azure.datafactory.tools/0.101.0/private/Deploy-AdfObjectOnly.ps1:135 char:13�[0m
2023-03-29T08:41:42.3159873Z �[31;1m+ New-AzResource `�[0m
2023-03-29T08:41:42.3196750Z �[31;1m+ ~~~~~~~~~~~~~~~~�[0m
2023-03-29T08:41:42.3204598Z �[31;1m+ CategoryInfo : CloseError: (:) [New-AzResource], Exception�[0m
2023-03-29T08:41:42.3205922Z �[31;1m+ FullyQualifiedErrorId : Microsoft.Azure.Commands.ResourceManager.Cmdlets.Implementation.NewAzureResourceCmdlet�[0m
2023-03-29T08:41:42.4383179Z ##[error]Script failed with exit code: 1
Activity
NowinskiK commentedon Oct 16, 2026
Thanks @supercomfysoft, and sorry for the long wait. This is still valid in the current version: every object is deployed with a single API call and no retry. Since #472 (v1.17.1) a failed deployment stops the publish, so one temporary timeout fails the whole run.
Observation
In your log the request started at
08:40:02.16and failed at08:41:42.31, almost exactly 100 seconds later. That is the defaultHttpClienttimeout, so the time limit ran out on the client side, and ADF may well have finished the change anyway. Retrying is safe: these calls create or update an object (HTTPPUT), so running one twice gives the same result.Planned approach
BadRequest),NotFound, authorization and so on fail on the first attempt, as today.$opt.MaxRetries(default3;0turns retries off).AzResourcedeploy method). Starting and stopping triggers already has its own retry and stays as it is.MaxRetries = 0.Merge pull request #510 from Azure-Player/issue-295-deploy-retry