Products went missing, and Adobe wasn't why: auditing a render farm before replacing it
A company wanted to replace an expensive render farm. The audit showed products were lost to five bugs in their own code, which any new tool would have inherited.
Context. An e-commerce company prints personalized merchandise. Every design is rendered on 21 Windows servers running old Adobe versions that can no longer be downloaded. That is ~$6.7k a month and up to 630k jobs. New products could wait 8.5 hours in the queue, and some never reached the store at all.
What I found. Five causes in the job runner, not in Adobe:
- A job leaves the queue before rendering, so a server crash loses it.
- Adobe errors are never caught or retried.
- A design counts as done as soon as one of its files exists.
- Nothing acts on the health-check status.
- A units bug means stuck jobs are never cleaned up.
Options. I compared five paths. Canva was out on its export cap: 5k a day against the 21k needed. Template SaaS tools cost more than the current servers. Adobe's cloud API only wins at a certain vendor price, and a custom Linux renderer needs a pilot.
Recommendation. Fix how the system runs before swapping the tool: stop losing jobs, a separate queue for new products, smaller servers. Then a pilot on ~10 templates with print-quality checks on 300 real jobs.