High-traffic SSR marketplace
A production marketplace whose server crashed every couple of hours. It was reported as a memory leak; it turned out to be something else. This is the part I worked on, as the frontend developer in the team.
- Engagement
- Since 2025, ongoing
- Role
- Frontend developer in the team
- Scale
- High-traffic production marketplace
The problem
The SSR server crashed roughly every two hours in production and had to be restarted. It was reported to me as a memory leak, and the issue did not show up locally.
The problems showed up in the frontend first, but the causes were not only there. I worked on it together with the rest of the team: backend and DevOps engineers and the CTO.
What I did
My part of the work:
Added logging to see what was happening
Added logs for requests and memory usage, in the frontend code and on the server, to get a clear picture of what was happening on the live platform.
Found the real cause: SSR on every request
The data showed it was not a memory leak. Every request, including bot traffic and irrelevant routes, triggered server-side rendering, on top of slow backend calls. I added an allowlist of pages that render on the server, with route validation, served everything else client-side, and capped concurrent server renders at six, falling back to client-side rendering above that.
Frontend side of a backend migration
Moved the frontend to a new backend architecture while the site kept serving users, with no downtime.
Ongoing frontend work
I still work on the product, building new features and keeping the existing ones stable.
The outcome
What changed:
- The crashes stopped. The fix was verified together with the DevOps team and the CEO.
- The migration to the new backend went through with no downtime.
Stack
React with server-side rendering, and Node.js. The problems showed up in the frontend, but finding them meant looking at the whole platform, not just one layer.
Does this sound like your app?
If your app has a similar problem, an audit is a low-risk way to start: read-only access, a written report and a prioritized list of fixes.