[JENKINS-49077] pipeline checkout step does not perform retries

Regular AbstractProject builds have a loop for retrying SCM checkouts when they fail throwing IO Exceptions. This retry loop is useful, because in certain circumstances the git server may disconnect, such as the following exception:

 

ERROR: Error fetching remote repo ''

hudson.plugins.git.GitException: Failed to fetch

at hudson.plugins.git.GitSCM.fetchFrom(GitSCM.java:812)

at hudson.plugins.git.GitSCM.retrieveChanges(GitSCM.java:1079)

at hudson.plugins.git.GitSCM.checkout(GitSCM.java:1110)

at org.jenkinsci.plugins.workflow.steps.scm.SCMStep.checkout(SCMStep.java:113)

at org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:130)

at org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:59)

at org.jenkinsci.plugins.workflow.job.WorkflowRun.run(WorkflowRun.java:262)

at hudson.model.ResourceController.execute(ResourceController.java:97)

at hudson.model.Executor.run(Executor.java:405)

Caused by: hudson.plugins.git.GitException: Command "git fetch --tags --progress " returned status code 128:

stdout:

stderr: key_load_public: invalid format

Received disconnect from port : Too many concurrent connections

Connection to closed by remote host.

fatal: Could not read from remote repository.

Basically, sometimes this server gets too much load and decides to disconnect the clones. Unfortunately this behavior can't be fixed on that end, and thus the retry loop is the only way to catch this.

In regular pipelines, I can write pipeline code which does the retry for myself. However, I'm unable to do this for the checkout step that occurs as part of fetching the pipeline code.

Is it possible to add the retry logic to the checkout step so that it will attempt to retry the checkout when it fails under this disconnection?

---
Originally reported by jekeller, imported from: pipeline checkout step does not perform retries

• status: Open
  
• priority: Major
  
• component(s): workflow-cps-plugin, workflow-scm-step-plugin
  
• resolution: Unresolved
  
• votes: 7
  
• watchers: 7
  
• imported: 2025-12-07
  

Regular AbstractProject builds have a loop for retrying SCM checkouts when they fail throwing IO Exceptions. This retry loop is useful, because in certain circumstances the git server may disconnect, such as the following exception:


 

ERROR: Error fetching remote repo '<REMOTE>'

hudson.plugins.git.GitException: Failed to fetch <URL>

	at hudson.plugins.git.GitSCM.fetchFrom(GitSCM.java:812)

	at hudson.plugins.git.GitSCM.retrieveChanges(GitSCM.java:1079)

	at hudson.plugins.git.GitSCM.checkout(GitSCM.java:1110)

	at org.jenkinsci.plugins.workflow.steps.scm.SCMStep.checkout(SCMStep.java:113)

	at org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:130)

	at org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:59)

	at org.jenkinsci.plugins.workflow.job.WorkflowRun.run(WorkflowRun.java:262)

	at hudson.model.ResourceController.execute(ResourceController.java:97)

	at hudson.model.Executor.run(Executor.java:405)

Caused by: hudson.plugins.git.GitException: Command "git fetch --tags --progress <URL> <REFSPEC>" returned status code 128:

stdout: 

stderr: key_load_public: invalid format

Received disconnect from <SERVER> port <PORT>: Too many concurrent connections

Connection to <SERVER> closed by remote host.

fatal: Could not read from remote repository.

Basically, sometimes this server gets too much load and decides to disconnect the clones. Unfortunately this behavior can't be fixed on that end, and thus the retry loop is the only way to catch this.


In regular pipelines, I can write pipeline code which does the retry for myself. However, I'm unable to do this for the checkout step that occurs as part of fetching the pipeline code.


Is it possible to add the retry logic to the checkout step so that it will attempt to retry the checkout when it fails under this disconnection?


 

environment

```
Jenkins 2.60.1

Pipeline version uncertain
```

Trace the checkout path from org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create through org.jenkinsci.plugins.workflow.steps.scm.SCMStep.checkout, using the reported GitException as the failure case. Determine how regular AbstractProject builds retry IO failures and identify the relevant workflow component tests. Done means the pipeline-code checkout retries after the described server disconnection instead of failing immediately.

Written by the indexing model from the issue text.