Turns out, you cannot use a key-pair with a passphrase in this situation. 
SSH key-pairs without a passphrase works correctly (but passphrases are 
mandatory in our shop).

On Monday, 20 July 2020 14:14:47 UTC-4, Randall Becker wrote:
>
> I'm going to have to dig deeper and probably debug the GitSCM plugin on 
> the agent. -Dibm.file.encoding does not help the situation. I have a call 
> later today that might shed some light on the situation.
>
> On Friday, 17 July 2020 09:47:19 UTC-4, Mark Waite wrote:
>>
>>
>> https://github.com/jenkinsci/git-client-plugin/blob/eeec334af0b6447f3db9fb88d55728911a092d73/src/main/java/org/jenkinsci/plugins/gitclient/CliGitAPIImpl.java#L2411
>>  has 
>> specific code for z/OS.  I do not have access to z/OS so cannot test that 
>> code.  It was merged based on a submission from people that said it works.  
>>
>> You're welcome to suggest improvements in that area with the 
>> understanding that I can only evaluate that code by inspection, not by 
>> execution.
>>
>> Mark Waite
>>
>> On Thu, Jul 16, 2020 at 7:49 AM Randall Becker <[email protected]> wrote:
>>
>>> That's what we were trying to do originally. There is a problem. When 
>>> GitSCM creates the GIT_SSH content on z/OS agent the file name is encoded 
>>> as an IBM1047 EBCDIC regardless of the -Dfile.encoding argument, even 
>>> though the original private key and passphrase are coming from a 
>>> UTF8/US-ASCII controller. When this goes to git and then SSH, the file is 
>>> still encoded as IBM1047 and is when it hits the KEX code, fails. When we 
>>> use the SSH Agent, this problem does not occur. I want to use the correct 
>>> using GitSCM credential ID, but it does not work. I do not have a decent 
>>> debug environment that would clearly demonstrate this, isolate the section 
>>> of code where this is (not) happening, or allow me to easily fix this. The 
>>> most important bit is that the private key should be encoded in UTF8 or 
>>> US-ASCII when supplied to GIT_SSH, not the default encoding. Somehow, the 
>>> SSH Agent Plugin does this correctly.
>>>
>>> On Monday, 13 July 2020 17:41:07 UTC-4, Mark Waite wrote:
>>>>
>>>> If the operation you're performing is a checkout, why use the ssh-agent 
>>>> wrapper?  Why not use the same credentials ID as an argument to checkout 
>>>> rather than wrapping checkout in ssh-agent?
>>>>
>>>> On Mon, Jul 13, 2020 at 8:45 AM Randall Becker <[email protected]> 
>>>> wrote:
>>>>
>>>>> I wish it was that simple. The issue definitely appears to be the 
>>>>> encoding of the private key during a key exchange. When using SSH-Agent 
>>>>> and 
>>>>> git commands from within a shell in the pipeline, the authentication 
>>>>> works 
>>>>> fine. So this is likely an interaction with the GitSCM plugin not being 
>>>>> aware of IBM-1047 encodings.
>>>>>
>>>>> On Sunday, 12 July 2020 16:08:31 UTC-4, Ivan Fernandez Calvo wrote:
>>>>>>
>>>>>> I think that this is the reason why it does not work 
>>>>>> https://support.cloudbees.com/hc/en-us/articles/224910467-Why-am-I-unable-to-authenticate-via-sshagent-inside-docker-
>>>>>>
>>>>>> El sábado, 11 de julio de 2020, 22:25:08 (UTC+2), Randall Becker 
>>>>>> escribió:
>>>>>>>
>>>>>>> I'm having issues trying to get an agent to authenticate using the 
>>>>>>> SSH Agent plugin on a R2.4 z/OS USS agent with a Docker Jenkins 
>>>>>>> controller. 
>>>>>>> The goal is to convince GitSCM to actually fetch properly. We get SSH 
>>>>>>> authentication errors no matter what happens. This is using Pipelines.
>>>>>>>
>>>>>>> I've tried 
>>>>>>>                         sshagent (credentials: ['mvs-randall']) {
>>>>>>>                             checkout([$class: 'GitSCM',
>>>>>>>                                 branches: [[name: '*/development']],
>>>>>>>                                 extensions: [
>>>>>>>                                     [$class: 'CleanBeforeCheckout'],
>>>>>>>                                     [$class: 'SubmoduleOption', 
>>>>>>> disableSubmodules: false, parentCredentials: true,
>>>>>>>                                         recursiveSubmodules: true, 
>>>>>>> reference: '', trackingSubmodules: false]],
>>>>>>>                                     
>>>>>>> doGenerateSubmoduleConfigurations: false, submoduleCfg: [],
>>>>>>>                                 userRemoteConfigs: [[url: 
>>>>>>> '[email protected]:proj/repo.git'']]])
>>>>>>>                         }
>>>>>>> and
>>>>>>>                         checkout([$class: 'GitSCM',
>>>>>>>                             branches: [[name: '*/development']],
>>>>>>>                             extensions: [
>>>>>>>                                 [$class: 'CleanBeforeCheckout'],
>>>>>>>                                 [$class: 'SubmoduleOption', 
>>>>>>> disableSubmodules: false, parentCredentials: true,
>>>>>>>                                     recursiveSubmodules: true, 
>>>>>>> reference: '', trackingSubmodules: false]],
>>>>>>>                                 doGenerateSubmoduleConfigurations: 
>>>>>>> false, submoduleCfg: [],
>>>>>>>                             userRemoteConfigs: [[credentialsId: 
>>>>>>> 'mvs-randall',url: '[email protected]:proj/repo.git']]])
>>>>>>>
>>>>>>> Both result in Permission denied (publickey).
>>>>>>>
>>>>>>> I've done the same thing on many other platforms with no problem. 
>>>>>>> This seems very R2.4 specific. There was a change in the supported file 
>>>>>>> encodings as well - we used to use -Dfile.encoding=utf8 in the agent 
>>>>>>> config 
>>>>>>> (because this is an IBM that likes EBCDIC), but had to move to 
>>>>>>> -Dfile.encoding=ISO8859-1 and everything seems messed up now. IBM had 
>>>>>>> this 
>>>>>>> funky script they recommend that massages the key into an IBM-1047 
>>>>>>> encoding 
>>>>>>> but that does not help at all - in fact the GitSCM agent cannot process 
>>>>>>> any 
>>>>>>> results if that script is used.
>>>>>>>
>>>>>>> Help! 
>>>>>>>
>>>>>>> TIA,
>>>>>>> Randall
>>>>>>>
>>>>>> -- 
>>>>> You received this message because you are subscribed to the Google 
>>>>> Groups "Jenkins Users" group.
>>>>> To unsubscribe from this group and stop receiving emails from it, send 
>>>>> an email to [email protected].
>>>>> To view this discussion on the web visit 
>>>>> https://groups.google.com/d/msgid/jenkinsci-users/1ece555d-921d-4a66-ba9d-2afe1cf212fao%40googlegroups.com
>>>>>  
>>>>> <https://groups.google.com/d/msgid/jenkinsci-users/1ece555d-921d-4a66-ba9d-2afe1cf212fao%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>>> .
>>>>>
>>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "Jenkins Users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to [email protected].
>>> To view this discussion on the web visit 
>>> https://groups.google.com/d/msgid/jenkinsci-users/b071384e-8160-4522-9237-d036af9a6c15o%40googlegroups.com
>>>  
>>> <https://groups.google.com/d/msgid/jenkinsci-users/b071384e-8160-4522-9237-d036af9a6c15o%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>

-- 
You received this message because you are subscribed to the Google Groups 
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jenkinsci-users/d10d3393-1dcb-4673-9ca5-0817ea7630b9o%40googlegroups.com.

Reply via email to