WIP: Fix Vertex AI scheduler 64-character limit for long pipeline names (#2262)

[hrushi2602-cyber]

Description

Relates to Issue: #2262

The Problem

When submitting a pipeline with a name longer than 64 characters to the Vertex AI orchestrator, the API rejects the payload due to strict display_name character limits. Simply slicing the string at 64 characters risks name collisions if multiple pipelines share a long prefix.

The Solution

Introduced a local helper function, sanitize_vertex_job_name, inside vertex_orchestrator.py that intercepts both orchestrator_run_name and pipeline_name.

• If the name is 64 characters or fewer, it passes through unchanged.
• If it exceeds 64 characters, it generates an MD5 hash of the full name, extracts an 8-character short hash, and appends it to the truncated string.
• This ensures the final display_name is exactly 64 characters and guaranteed unique to avoid overwriting jobs.

Current Status (WIP)

I am opening this as a Draft PR to show intent and verify this hashing approach aligns with the maintainers' expectations before wiring up the full test coverage.

• Implement hash-truncation logic
• Apply to Vertex AI payload variables
• Add unit tests (Pending feedback on the approach)

Looking forward to any feedback!

[CLAassistant]

All committers have signed the CLA.

[schustmi]

Hey @hrushi2602-cyber, thanks for the PR! Can you please rebase your changes on the develop branch and also develop as a base branch for your PR.

Regarding the implementation:
We already have something very similar to this implemented in the get_orchestrator_run_name(...) function (see src/zenml/orchestrators/utils.py). In my opinion, it would be nicest if we do the following:

• Add a new function in the string_utils module. This should be generic, receives as input some original string, a optional maximum length, a suffix length and maybe also a separator string (e.g. "-" or "_") . It should then generate a random suffix, truncate the original string so a suffix of that length fits and then merge them.
• We call this function both from get_orchestrator_run_name(...) as well as the one you introduced (sanitize_vertex_job_name(...))

[hrushi2602-cyber]

Thanks for the feedback! I've updated the PR based on your suggestions. Here is a summary of the changes:

Rebased and Targeted develop: I have successfully changed the PR base branch to develop and rebased my local commits on top of the latest changes from the upstream develop branch.

Centralized String Utility: Added a new generic function append_random_suffix in src/zenml/utils/string_utils.py. It handles suffix generation, strict length limit enforcement, and safe string merging. I also included a charset argument to ensure backward compatibility with existing components that might expect hexadecimal strings.

Refactored get_orchestrator_run_name: Updated src/zenml/orchestrators/utils.py to offload the string truncation and merging logic to the new utility. The original logic that dynamically shrinks the suffix (while enforcing the minimum of 8 characters) and uses the hexadecimal charset was fully preserved so this won't break downstream orchestrators.

Refactored sanitize_vertex_job_name: This function now acts as a pre-filter to enforce Vertex-specific rules (lowercase, alphanumeric, starting with a letter, replacing underscores with hyphens), and then defers to append_random_suffix to safely enforce the 64-character limit.

[schustmi]

Few minor things, other than that looks good to me

[schustmi]

I think we should also use separator="" here for backwards compatibility. All orchestrators use this function, and they might not allow underscores.

[schustmi]

Suggested change

# Use the provided charset instead of hardcoding alphanumeric

[schustmi]

Please adjust these to follow the google docstring format.

[schustmi]

Suggested change

	charset="0123456789abcdef" # <--- Ensures 100% backward compatibility
	charset="0123456789abcdef"

[hrushi2602-cyber]

All requested styling changes have been made! Here is what was updated in the latest commits:

Google Docstrings: Updated the append_random_suffix utility to strictly follow the Google docstring format.

Linter Fixes: Swapped single quotes to double quotes, removed the duplicate re import, and properly sorted the import block in vertex_orchestrator.py.

I ran the ruff checks locally on the files changed by me and everything is passing clean with Ruff now. Let me know if you need anything else.

[hrushi2602-cyber]

@schustmi Just a gentle reminder! I just applied your suggested changes just check once and let me know... Just commenting in case it got burried under due to the weekend!!