Alright, gather ‘round the campfire of fiscal regret and let me tell you a tale of digital hoarding and its direct debit from your wallet. For months, I’d glance at my team’s CI/CD platform invoice with a vague sense of unease, like a background app draining your phone battery. We were hovering around a **$650/month** burn on a popular managed platform. Not small-team-change money.
The usual suspects were all present: parallel jobs, hefty runners for Docker builds, the occasional midnight test suite orgy. But I had a hunch the ghost in the machine wasn't just live compute. So I went digging into the storage and retention policies—or, more accurately, the *lack thereof*.
Turns out, our default settings were preserving **EVERY** workflow run log and its associated artifacts (test results, coverage reports, bundled packages, Docker layers)… **forever**. Or at least, since we migrated to this platform two years ago. We had terabytes of historical data just… sitting there. Charged at that premium “convenience” storage rate.
Here’s the brutal breakdown of the cleanup operation:
* **Workflow run logs:** Default was "Keep all." Set a retention policy of 90 days. The platform doesn't charge *directly* for the log storage per se in our plan, but it counts towards overall account data bloat. More importantly, it cleans the UI.
* **Artifacts:** This was the killer. Every pipeline generated 3-5 artifacts, some up to 2GB each (hello, node_modules caches and full Docker context tarballs). Retention? **None.** Manually set to expire after 30 days.
* **Container registry:** Don't get me started. We had 14-month-old ephemeral test images still lurking. Pruned anything older than 30 days that wasn't tagged for production or staging.
The **immediate result** after the policies propagated and the cleanup scripts ran? Our next monthly invoice was **$447**. A **$203** saving. For. Doing. Literally. Nothing. Except deleting digital dust.
The takeaway isn't that you should start deleting everything—it's that **retention is a cost variable**. Managed CI/CD is fantastic until you realize its business model thrives on your inattention to data lifecycle. It's like paying for a storage unit full of boxes you'll never open again.
My question to the room: What are your artifact/log retention policies? Has anyone found a sweet spot between "useful for debugging" and "financially irresponsible"? And does anyone actually *trust* their platform's built-in cleanup tools, or do you run external cron jobs for this?
chloe
Demos are just theater. Show me the real workflow.