Ask HN: Why is HN so often down?
Frequent error messages and brief outages on Hacker News are prompting users to question why a text-only site struggles with reliability while serving a large, highly technical audience. Commenters point to a single old, single-core server, non-autoscaling infrastructure, and aggressive anti-bot rate limiting as key factors, compounded by a recent spike in traffic from higher Google search rankings. Many accept the trade-off as a consequence of HN’s minimalist, non-commercial operation run by a very small team, though some are surprised such a central industry hub tolerates relatively low uptime in 2024.
Perceived Frequency and Patterns of Downtime
- Experiences vary widely:
- Some daily users over many years report almost never seeing errors.
- Others say outages or 5xx errors have become frequent in the last months, even multiple times per day.
- Some report issues only this week; others say it’s been months.
- Problems can be intermittent: a few minutes of errors, then fine.
- Several note errors more often during certain daytime hours (e.g., mid-morning to early afternoon CET).
Infrastructure, Architecture, and Capacity
- HN is described as running on a single, old, one‑core server, likely without autoscaling or replicas.
- Codebase is old and largely single-threaded, written in Arc (on Racket), which may make scaling and refactoring harder.
- Some argue this is “the easiest kind of app” to cache and shard; others suggest the legacy stack and small team make big changes slow.
- A 500 error page is said to indicate backend unavailability or slow restarts, not just rate limiting.
Rate Limiting, Anti‑Bot Measures, and Login State
- Users frequently hit strict rate limits when voting, commenting, or favoriting, especially when logged out.
- There is an anti‑bot system that can temporarily ban IPs, particularly for high activity or shared IPs (VPNs, corporate networks).
- Logged‑in browsing can reduce the chance of anti‑bot bans (per past moderator comments), but:
- Logged‑out users benefit more from cached static pages under heavy load.
- Some users suspect “shadow bans” or content‑based throttling; others explicitly dismiss this as unfounded.
Traffic Growth and External Factors
- Several speculate that recent Google ranking changes have significantly increased traffic, stressing the infrastructure.
- Some consider possible network issues (e.g., cable cuts), but this is not confirmed.
Philosophy, Expectations, and Alternatives
- Many emphasize HN is a free, mostly ad‑free service run by a very small team; high “number of nines” availability is not a goal.
- Some find the occasional outages acceptable or even “charming”; others are frustrated by wasted time.
- Suggestions range from serverless/SaaS migrations to better caching; these ideas are met with skepticism and jokes.
- A few users browse via third‑party search frontends (e.g., hn.algolia.com) for better uptime.