Understand the behavior of infinity upgrade remediation

[mo-saeed]

Hello,

I would like to achieve this block from flux v1 configuration

 rollback:
    enable: true
    retry: true
    maxRetries: 0

as per my understanding this can do it

 upgrade:
   remediation:
     retries: -1
 install:
   remediation:
     retries: -1

but i am bit not sure how this works exactly, is it one of these scenarios

• first upgrade attempt failed
• rollback
• rollback success
• upgrade again
• fail
• rollback
• rollback success

or

• first upgrade attempt failed
• rollback
• rollback success
• done

what i want is sth like this, which was done in flux v1

• first upgrade attempt failed
• rollback
• rollback failed
• rollback again
• rollback success
• done

thanks alot in advance

[mo-saeed]

Hi @stefanprodan Hi @hiddeco, maybe you have a quick answer, I would appreciate it so i can move on with the migration.

Thanks

[kingdonb]

Have I understood correctly (I'll try to reiterate what I think you're asking for)

When a release fails, it should not retry, it should roll back. In the event that rollback fails, it should also rollback again, as many times as needed until the rollback succeeds. Is that correct?

I did not know or think that rollbacks could fail, or be retried by Helm. I don't think the upstream Helm CLI retries anything, so it wouldn't depend on upstream support, (retries are in Flux's domain exclusively)