https://prefect.io logo
Join Slack
Powered by
Hi <@ULVA73B9P>, I’m having an issue with the Wind...
# ask-marvin
m
Hi @Marvin, I’m having an issue with the Windows VM where my Prefect worker is running: from time to time it loses connection, and as a result my flows fail to start with the following error:
Copy code
RuntimeError: Failed to clone repository '<https://github.com/prometeia-wam/da-data-pipeline-customer.git>' with exit code 128.
To work around this, I’ve been looking into adjusting my 
prefect.yaml
. Here’s the relevant part of the file:
Copy code
# Generic metadata about this project
name: prefect-xxx

# pull section allows you to provide instructions for cloning this project in remote locations
pull:
- prefect.deployments.steps.git_clone:
    repository: <https://github.com/prometeia-wam/da-data-pipeline-customer.git>
    branch: "{{ prefect.variables.git_branch }}"
    access_token: "{{ prefect.blocks.secret.github-tech-user-token }}"

deployments:
- name: xxxx-upload
  version:
  tags: ['CUSTOMER','BPER']
  description: Upload outputs for BPER
  schedule:
      cron: "30 22 * * 1-5"
      timezone: "Europe/Rome"
      active: "{{ prefect.variables.is_schedule_active }}"
  entrypoint: flows/assets_uploader.py:assets_uploader
  parameters: {asset_ids: ["xxxx-gpp_benchmark_description"]}
  work_pool:
    name: customer-workpool
    work_queue_name: default
    job_variables: {}
What I’d like to do is extend the 
pull
section: if the 
git_clone
step fails, I’d like to run a 
prefect.deployments.steps.run_shell_script
step that retries the clone. Do you have any suggestions on the best way to set this up in 
prefect.yaml
?
m
✅ thought for 182.2 seconds
Thanks for the detail — this is a common pain point when the worker has intermittent network hiccups. Short answer: in Prefect 2.x, pull steps don’t support retries or conditional/fallback logic. If a step fails, the pull sequence stops and the flow never starts. There’s an open feature request to add retries to 
git_clone
here: Add retries to pull step (git clone) Given that, the most reliable approach is to replace the 
git_clone
pull step with a 
run_shell_script
step that performs the clone with retry/backoff itself. On Windows, 
run_shell_script
uses 
cmd.exe
by default, so here’s a drop-in example that keeps your templating and token block but adds retries:
Copy code
pull:
  - prefect.deployments.steps.run_shell_script:
      # Keep secrets out of the script body; reference them via env
      env:
        GITHUB_TOKEN: "{{ prefect.blocks.secret.github-tech-user-token }}"
        GIT_BRANCH: "{{ prefect.variables.git_branch }}"
      expand_env_vars: true
      # Consider setting to false if you're worried about logs (Prefect will try to redact, but env is safest)
      stream_output: true
      script: |
        @echo off
        setlocal enabledelayedexpansion

        set REPO_OWNER=prometeia-wam
        set REPO_NAME=da-data-pipeline-customer
        set REPO_DIR=%REPO_NAME%
        set RETRIES=5
        set SLEEP=15

        echo Cloning %REPO_OWNER%/%REPO_NAME% branch %GIT_BRANCH% with up to %RETRIES% retries...

        for /L %%i in (1,1,%RETRIES%) do (
          if exist "%REPO_DIR%" rmdir /s /q "%REPO_DIR%"

          rem Use a token-authenticated URL; avoid printing the URL to logs
          git clone --depth 1 --branch "%GIT_BRANCH%" "<https://%GITHUB_TOKEN>%@github.com/%REPO_OWNER%/%REPO_NAME%.git" "%REPO_DIR%"
          if !errorlevel! EQU 0 goto :success

          echo Clone failed (attempt %%i/%RETRIES%). Retrying in %SLEEP% seconds...
          timeout /t %SLEEP% /nobreak >nul
        )

        echo All retries failed.
        exit /b 1

        :success
        echo Clone succeeded.
Then remove the original 
prefect.deployments.steps.git_clone
entry. Everything else in your 
prefect.yaml
can stay the same. Notes and options: - Security: Passing the token via 
env
+ 
expand_env_vars: true
keeps it out of the script body. You can also set 
stream_output: false
if you want to be extra cautious about logging. - Performance: Using 
--depth 1
reduces data transfer and can make retries succeed faster. - PowerShell alternative: If you prefer PowerShell, I can share a PowerShell retry loop; we’d invoke it via 
run_shell_script
(either as a PowerShell one-liner or by calling 
powershell -Command
). - Docs for pull steps and run_shell_script: Pull steps docs and a reference to 
run_shell_script
under steps: Deployment guide If you’d rather keep using the built-in 
git_clone
, there isn’t a way to “fallback” on failure today. The alternative pattern to avoid runtime git entirely is to package code with your runtime (e.g., ship a Docker image or pull from object storage), which sidesteps intermittent git network issues — happy to outline that if helpful. Would you like the PowerShell version of the script, or are you good with the cmd-style loop above?
m
@Marvin questa è la configurazione che trovo a front end nel deployment: [ { "prefect.deployments.steps.run_shell_script": { "env": { "GIT_BRANCH": "{{ prefect.variables.git_branch }}", "GITHUB_TOKEN": "{{ prefect.blocks.secret.github-tech-user-token }}" }, "script": "@echo off\nsetlocal enabledelayedexpansion\n\nset REPO_OWNER=prometeia-wam\nset REPO_NAME=da-data-pipeline-customer\nset REPO_DIR=%REPO_NAME%\nset RETRIES=5\nset SLEEP=15\n\necho Cloning %REPO_OWNER%/%REPO_NAME% branch %GIT_BRANCH% with up to %RETRIES% retries...\n\nfor /L %%i in (1,1,%RETRIES%) do (\n if exist \"%REPO_DIR%\" rmdir /s /q \"%REPO_DIR%\"\n\n rem Use a token-authenticated URL; avoid printing the URL to logs\n git clone --depth 1 --branch \"%GIT_BRANCH%\" \"https://%GITHUB_TOKEN%@github.com/%REPO_OWNER%/%REPO_NAME%.git\" \"%REPO_DIR%\"\n if !errorlevel! EQU 0 goto :success\n\n echo Clone failed (attempt %%i/%RETRIES%). Retrying in %SLEEP% seconds...\n timeout /t %SLEEP% /nobreak >nul\n)\n\necho All retries failed.\nexit /b 1\n\n:success\necho Clone succeeded.\n", "stream_output": true, "expand_env_vars": true } } ] mi da l'errore: Flow could not be retrieved from deployment. Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 154, in run_steps step_output = await run_step(step, upstream_outputs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 125, in run_step result = await from_async.call_soon_in_new_thread( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 327, in aresult return await asyncio.wrap_future(self.future) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 389, in _run_async result = await coro ^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\utility.py", line 192, in run_shell_script raise RuntimeError( RuntimeError: 
run_shell_script
failed with error code 1: %%i was unexpected at this time. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\engine.py", line 420, in retrieve_flow_then_begin_flow_run else await load_flow_from_flow_run(flow_run, client=client) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\client\utilities.py", line 78, in with_injected_client return await fn(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\deployments.py", line 301, in load_flow_from_flow_run output = await run_steps(deployment.pull_steps) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 182, in run_steps raise StepExecutionError(f"Encountered error while running {fqn}") from exc prefect.deployments.steps.core.StepExecutionError: Encountered error while running prefect.deployments.steps.run_shell_script
m
✅ thought for 96.6 seconds
Grazie per il dettaglio — qui ci sono tre problemi che spiegano l’errore “%%i was unexpected at this time.” e il fallimento del clone: 1) In Windows 
cmd.exe
, quando esegui uno script “inline” (come fa 
run_shell_script
), la variabile del 
for
deve usare una sola percentuale (
%i
), non 
%%i
(le doppie percentuali servono solo dentro file .bat). 2) 
expand_env_vars: true
fa espandere prima del run i placeholder 
%VAR%
con l’ambiente del processo Python (non con 
env:
che imposti tu). Questo può trasformare 
(%RETRIES%)
in vuoto e rompere il 
for
. Meglio disattivarlo. 3) La stringa dell’URL è stata “contaminata” dal front-end/Slack: vedo 
https://%<mailto:GITHUB_TOKEN%@github.com|GITHUB_TOKEN%@github.com>/...
— va sostituita con 
<https://%GITHUB_TOKEN>%@github.com/...
. Prova questa configurazione corretta (CMD), con retry/backoff e senza espansione preventiva:
Copy code
pull:
  - prefect.deployments.steps.run_shell_script:
      env:
        GIT_BRANCH: "{{ prefect.variables.git_branch }}"
        GITHUB_TOKEN: "{{ prefect.blocks.secret.github-tech-user-token }}"
      expand_env_vars: false
      stream_output: true
      script: |
        @echo off
        setlocal EnableDelayedExpansion

        set REPO_OWNER=prometeia-wam
        set REPO_NAME=da-data-pipeline-customer
        set REPO_DIR=%REPO_NAME%
        set RETRIES=5
        set SLEEP=15

        echo Cloning %REPO_OWNER%/%REPO_NAME% branch %GIT_BRANCH% with up to %RETRIES% retries...

        for /L %i in (1,1,%RETRIES%) do (
          if exist "%REPO_DIR%" rmdir /s /q "%REPO_DIR%"

          rem Token nel campo utente dell'URL (attenzione ai log)
          git clone --depth 1 --branch "%GIT_BRANCH%" "<https://%GITHUB_TOKEN>%@github.com/%REPO_OWNER%/%REPO_NAME%.git" "%REPO_DIR%"
          if !errorlevel! EQU 0 goto :success

          echo Clone failed (attempt %i/%RETRIES%). Retrying in %SLEEP% seconds...
          timeout /t %SLEEP% /nobreak >nul
        )

        echo All retries failed.
        exit /b 1

        :success
        echo Clone succeeded.
Note importanti: - Usa 
%i
(non 
%%i
) nel 
for
. - 
expand_env_vars: false
per evitare che 
%RETRIES%
, 
%REPO_NAME%
, etc. vengano espansi prima che 
set
li definisca. - Correggi l’URL del clone per rimuovere il frammento 
mailto
che vedi nel front end. - Se vuoi minimizzare il rischio che il token compaia nei log, metti 
stream_output: false
. In alternativa, puoi evitare il token nell’URL e usare un header (attenzione: il token sarà comunque visibile nella command line del processo):
Copy code
git -c http.extraHeader="Authorization: Bearer %GITHUB_TOKEN%" clone --depth 1 --branch "%GIT_BRANCH%" "<https://github.com/%REPO_OWNER%/%REPO_NAME%.git>" "%REPO_DIR%"
Se preferisci PowerShell (meno insidie con le percentuali), posso fornirti anche la versione PS con retry. Vuoi che te la condivida? Inoltre, qui trovi la feature request per aggiungere retry al pull step 
git_clone
nativo: Add retries to pull step (git clone).
m
@Marvin ora mi da l'errore --> Flow could not be retrieved from deployment. Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 154, in run_steps step_output = await run_step(step, upstream_outputs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 125, in run_step result = await from_async.call_soon_in_new_thread( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 327, in aresult return await asyncio.wrap_future(self.future) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 389, in _run_async result = await coro ^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\utility.py", line 192, in run_shell_script raise RuntimeError( RuntimeError: 
run_shell_script
failed with error code 128: Cloning into '%REPO_DIR%'... fatal: unable to access 'https://github.com/%REPO_OWNER%/%REPO_NAME%.git/': The requested URL returned error: 400 The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\engine.py", line 420, in retrieve_flow_then_begin_flow_run else await load_flow_from_flow_run(flow_run, client=client) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\client\utilities.py", line 78, in with_injected_client return await fn(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\deployments.py", line 301, in load_flow_from_flow_run output = await run_steps(deployment.pull_steps) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 182, in run_steps raise StepExecutionError(f"Encountered error while running {fqn}") from exc prefect.deployments.steps.core.StepExecutionError: Encountered error while running prefect.deployments.steps.run_shell_script --> questa la configurazione che vedo a front end -->
Copy code
[
  {
    "prefect.deployments.steps.run_shell_script": {
      "env": {
        "GIT_BRANCH": "{{ prefect.variables.git_branch }}",
        "GITHUB_TOKEN": "{{ prefect.blocks.secret.github-tech-user-token }}"
      },
      "script": "@echo off\nsetlocal EnableDelayedExpansion\n\nset REPO_OWNER=prometeia-wam\nset REPO_NAME=da-data-pipeline-customer\nset REPO_DIR=%REPO_NAME%\nset RETRIES=5\nset SLEEP=15\n\necho Cloning %REPO_OWNER%/%REPO_NAME% branch %GIT_BRANCH% with up to %RETRIES% retries...\n\nfor /L %i in (1,1,%RETRIES%) do (\n  if exist \"%REPO_DIR%\" rmdir /s /q \"%REPO_DIR%\"\n\n  rem Token nel campo utente dell'URL (attenzione ai log)\n  git clone --depth 1 --branch \"%GIT_BRANCH%\" \"<https://%GITHUB_TOKEN>%@github.com/%REPO_OWNER%/%REPO_NAME%.git\" \"%REPO_DIR%\"\n  if !errorlevel! EQU 0 goto :success\n\n  echo Clone failed (attempt %i/%RETRIES%). Retrying in %SLEEP% seconds...\n  timeout /t %SLEEP% /nobreak >nul\n)\n\necho All retries failed.\nexit /b 1\n\n:success\necho Clone succeeded.\n",
      "stream_output": true,
      "expand_env_vars": false
    }
  }
]
m
✅ thought for 117.3 seconds
Grazie, ora è chiaro cosa sta succedendo. Perché ora fallisce con 400 e vede letterali come `%REPO_DIR%`: - Su Windows, 
run_shell_script
esegue il contenuto con PowerShell. In PowerShell le variabili non sono nel formato 
%VAR%
, quindi 
%REPO_DIR%
, 
%REPO_OWNER%
, ecc. non vengono espanse e finiscono nella URL letteralmente. - Inoltre, nel front-end vedo ancora i chevron “< >” attorno alla URL (
"<https://%GITHUB_TOKEN>%@github.com/..."
): quelli vengono dal rendering e rompono la URL. Soluzione più semplice: usa una versione PowerShell del retry e passa il token tramite header HTTP (così eviti anche la URL con credenziali). Ecco una configurazione pronta all’uso:
Copy code
pull:
  - prefect.deployments.steps.run_shell_script:
      env:
        GIT_BRANCH: "{{ prefect.variables.git_branch }}"
        GITHUB_TOKEN: "{{ prefect.blocks.secret.github-tech-user-token }}"
      expand_env_vars: false
      stream_output: true
      script: |
        $ErrorActionPreference = 'Stop'

        $REPO_OWNER = 'prometeia-wam'
        $REPO_NAME  = 'da-data-pipeline-customer'
        $REPO_DIR   = $REPO_NAME
        $RETRIES    = 5
        $SLEEP      = 15

        Write-Host "Cloning $REPO_OWNER/$REPO_NAME branch $env:GIT_BRANCH with up to $RETRIES retries..."

        for ($i = 1; $i -le $RETRIES; $i++) {
          if (Test-Path $REPO_DIR) { Remove-Item -Recurse -Force $REPO_DIR }

          # Autenticazione via header: più sicura e senza problemi di quoting della URL
          git -c http.extraHeader="Authorization: Bearer $env:GITHUB_TOKEN" clone --depth 1 --branch "$env:GIT_BRANCH" "<https://github.com/$REPO_OWNER/$REPO_NAME.git>" "$REPO_DIR"
          if ($LASTEXITCODE -eq 0) {
            Write-Host "Clone succeeded."
            exit 0
          }

          Write-Host "Clone failed (attempt $i/$RETRIES). Retrying in $SLEEP seconds..."
          Start-Sleep -Seconds $SLEEP
        }

        Write-Error "All retries failed."
        exit 1
Note importanti: - Niente “< >” attorno alle URL nel 
script
. Devono essere stringhe normali. - Mantieni `expand_env_vars: false`: in questo modo la sostituzione di 
$env:GIT_BRANCH
e 
$env:GITHUB_TOKEN
avviene in PowerShell (non prima). - Se vuoi minimizzare la possibilità che il token finisca nei log, puoi anche impostare 
stream_output: false
. Alternativa: forzare CMD. Se preferisci mantenere la versione CMD, allora richiama esplicitamente CMD da PowerShell, ma è più fragile. La versione PowerShell sopra è la più pulita su Windows. Riferimento alla richiesta di feature per retry nativi del pull step `git_clone`: Add retries to pull step (git clone) Prova questa configurazione e dimmi se vedi ancora errori; in caso, condividi l’output “Clone failed …” della prima iterazione (senza token) così capiamo se è un problema di rete/proxy o di permessi del token su quel repo.
m
@Marvin ora dice: Flow could not be retrieved from deployment. Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 154, in run_steps step_output = await run_step(step, upstream_outputs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 125, in run_step result = await from_async.call_soon_in_new_thread( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 327, in aresult return await asyncio.wrap_future(self.future) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\_internal\concurrency\calls.py", line 389, in _run_async result = await coro ^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\utility.py", line 192, in run_shell_script raise RuntimeError( RuntimeError: 
run_shell_script
failed with error code 1: '$ErrorActionPreference' is not recognized as an internal or external command, operable program or batch file. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\engine.py", line 420, in retrieve_flow_then_begin_flow_run else await load_flow_from_flow_run(flow_run, client=client) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\client\utilities.py", line 78, in with_injected_client return await fn(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\deployments.py", line 301, in load_flow_from_flow_run output = await run_steps(deployment.pull_steps) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\PREFECT_GITHUB\venv\Lib\site-packages\prefect\deployments\steps\core.py", line 182, in run_steps raise StepExecutionError(f"Encountered error while running {fqn}") from exc prefect.deployments.steps.core.StepExecutionError: Encountered error while running prefect.deployments.steps.run_shell_script
m
✅ thought for 64.7 seconds
Perfetto, l’errore ora è chiaro: il contenuto PowerShell è stato eseguito in CMD. In Prefect 2.x su Windows, 
run_shell_script
usa CMD per default. Perciò: - Se vuoi usare PowerShell, devi invocarlo esplicitamente da CMD. - In alternativa, resta su CMD puro (più semplice) ma correggi URL/variabili. Ti propongo la soluzione più semplice: versione CMD con retry e autenticazione via header (evita di mettere il token nella URL e previene i 400 dovuti a encoding/quoting). Sostituisci il tuo 
pull
con questo (attenzione: niente “< >” attorno alle URL nel 
script
):
Copy code
pull:
  - prefect.deployments.steps.run_shell_script:
      env:
        GIT_BRANCH: "{{ prefect.variables.git_branch }}"
        GITHUB_TOKEN: "{{ prefect.blocks.secret.github-tech-user-token }}"
      expand_env_vars: false
      stream_output: true
      script: |
        @echo off
        setlocal EnableDelayedExpansion

        set REPO_OWNER=prometeia-wam
        set REPO_NAME=da-data-pipeline-customer
        set REPO_DIR=%REPO_NAME%
        set RETRIES=5
        set SLEEP=15

        echo Cloning %REPO_OWNER%/%REPO_NAME% branch %GIT_BRANCH% with up to %RETRIES% retries...

        for /L %i in (1,1,%RETRIES%) do (
          if exist "%REPO_DIR%" rmdir /s /q "%REPO_DIR%"

          rem Autenticazione via header (niente token nella URL)
          git -c http.extraHeader="Authorization: Bearer %GITHUB_TOKEN%" clone --depth 1 --branch "%GIT_BRANCH%" "<https://github.com/%REPO_OWNER%/%REPO_NAME%.git>" "%REPO_DIR%"
          if !errorlevel! EQU 0 goto :success

          echo Clone failed (attempt %i/%RETRIES%). Retrying in %SLEEP% seconds...
          timeout /t %SLEEP% /nobreak >nul
        )

        echo All retries failed.
        exit /b 1

        :success
        echo Clone succeeded.
Punti chiave: - Usa 
%i
(non 
%%i
) nel 
for
. - Mantieni 
expand_env_vars: false
(le variabili 
%...%
vengono espanse da CMD al momento giusto). - Evita assolutamente i chevron 
<>
nelle stringhe del `script`: quelli vengono da editor/rendering e rompono la URL. - Il token è passato via 
env
e viene usato nel header HTTP (più sicuro e robusto del token nella URL). - Verifica che 
git
sia nel PATH del worker. Se invece preferisci PowerShell, devi invocarlo esplicitamente da CMD. Ecco una variante che crea un file .ps1 temporaneo e lo esegue:
Copy code
pull:
  - prefect.deployments.steps.run_shell_script:
      env:
        GIT_BRANCH: "{{ prefect.variables.git_branch }}"
        GITHUB_TOKEN: "{{ prefect.blocks.secret.github-tech-user-token }}"
      expand_env_vars: false
      stream_output: true
      script: |
        @echo off
        setlocal
        set PS1=%TEMP%\prefect_clone.ps1
        >"%PS1%" (
          echo $ErrorActionPreference = 'Stop'
          echo $REPO_OWNER = 'prometeia-wam'
          echo $REPO_NAME  = 'da-data-pipeline-customer'
          echo $REPO_DIR   = $REPO_NAME
          echo $RETRIES    = 5
          echo $SLEEP      = 15
          echo Write-Host "Cloning $REPO_OWNER/$REPO_NAME branch $env:GIT_BRANCH with up to $RETRIES retries..."
          echo for ($i = 1; $i -le $RETRIES; $i++) {
          echo   if (Test-Path $REPO_DIR) { Remove-Item -Recurse -Force $REPO_DIR }
          echo   git -c http.extraHeader^="Authorization: Bearer $env:GITHUB_TOKEN" clone --depth 1 --branch "$env:GIT_BRANCH" "<https://github.com/$REPO_OWNER/$REPO_NAME.git>" "$REPO_DIR"
          echo   if ($LASTEXITCODE -eq 0) { Write-Host "Clone succeeded."; exit 0 }
          echo   Write-Host "Clone failed (attempt $i/$RETRIES). Retrying in $SLEEP seconds..."
          echo   Start-Sleep -Seconds $SLEEP
          echo }
          echo Write-Error "All retries failed."
          echo exit 1
        )
        powershell -NoProfile -NonInteractive -ExecutionPolicy Bypass -File "%PS1%"
        set EXITCODE=%ERRORLEVEL%
        del "%PS1%" 2>nul
        exit /b %EXITCODE%
Infine, ricordo che non esistono retry/fallback nativi per i pull steps in Prefect 2.x; c’è una richiesta aperta qui: Add retries to pull step (git clone) Se dopo questa modifica vedi ancora errori, incolla le prime 4-5 righe di output del pull step (senza token) e verifichiamo se è un tema di rete/proxy o permessi del token.
2 Views
Open in Slack
PreviousNext
Powered by