502 Bad Gateway
A 502 is one server tattling on another.
Bad Gateway means the front-door server (usually nginx or a CDN) asked the application server behind it for your page and got garbage or silence back. The finger-pointing is built into the error — the diagnosis is figuring out which server deserves it.
Diagnosis within one business day (Mon–Fri). Flat quote before any fix.
What’s actually happening
Modern hosting is a relay: a proxy or CDN out front, PHP or an app server behind it. A 502 means the handoff failed — the backend crashed, restarted, timed out hard, or answered in a way the front couldn't parse. Unlike a 500 (the app erred while running), a 502 often means the app never got to run at all.
That distinction decides ownership: intermittent 502s under load are usually the host's capacity or a crashing PHP process; constant 502s after a change are usually that change; 502s only through a CDN like Cloudflare, while the origin loads fine, point at the CDN handshake — SSL modes and timeouts between the layers.
The usual causes, ranked
After twenty-seven years of these calls, the odds are well mapped. Start at the top.
The backend process crashed or is restarting
PHP-FPM died, ran out of workers, or is cycling. The front door knocks; nobody's home. The classic intermittent 502.
Host capacity or a noisy neighbor
On shared hosting, the backend pool is shared too — someone else's traffic spike can 502 your site. Timing correlates with hours, not your changes.
CDN-to-origin handshake problems
Cloudflare or another CDN can't talk to your origin — SSL mode mismatches, origin timeouts, blocked IPs. The tell: site fine when you bypass the CDN.
A deploy or config change
New code, changed PHP versions, or a bad proxy config landed right before the 502s began. The timeline is the confession.
What you can safely try first
Nothing below can make things worse — that’s the selection criterion. Anything riskier belongs in professional hands, on a backup.
- 1
Refresh once, then note the pattern
Constant or intermittent? Everywhere or one page? 502s that come and go under load speak capacity; constant ones speak breakage. The pattern is half the diagnosis.
- 2
Check your host's status page
502s are the error most likely to be genuinely their outage. Two minutes of checking beats an hour of self-blame.
- 3
If you use Cloudflare, note whether it names itself
Cloudflare's branded 502 page means the CDN-origin handshake — a specific, fixable configuration conversation, not a mystery.
Stop and call when…
- 502s are intermittent under traffic and the host says all is well — capacity forensics with evidence is the fix
- It started after a deploy or PHP change and the rollback isn't obvious
- The CDN handshake needs SSL-mode surgery you haven't done before
From there it’s our job: a look within one business day, flat quote, and the $249 flat repair covers most cases of exactly this.
Site Down or White Screen — buy it now, skip the hunt.
500 and 502 errors, a white screen, a database connection error, a site that broke after a PHP update.
Back online first, then the cause. A full backup once it is up, and a note on what failed and how to keep it from happening again. We start within one business day (Monday–Friday) of payment. Full refund if we can't fix it.
Questions we hear a lot.
Is a 502 my fault or my host's?
Genuinely could be either — that's what makes it the finger-pointing error. Rough split: intermittent-under-load leans host; constant-after-a-change leans you (or whoever changed things); CDN-branded leans configuration between the layers. We diagnose with evidence so the right party fixes it — and we talk to hosts with logs in hand, which changes those conversations.
Is my site down during a 502?
Effectively yes for affected visitors — they get the error page, not your site. Intermittent 502s are sneakier: most visitors get through, some bounce off, and analytics quietly sags without an obvious outage. Worth fixing at down-site urgency either way.
What does the flat fix cover?
The diagnosis with evidence (whose layer, what mechanism), the fix where it's in your stack — config, PHP workers, CDN handshake — or the documented escalation to your host where it's theirs, pursued until resolved. $249, refunded if we can't get it sorted.
Related symptoms & help
Errors travel in packs. If this one visited, check its friends.
- 503 Service UnavailableThe sibling: server present but refusing.
- 504 Gateway TimeoutThe sibling: server present but too slow.
- 500 ErrorWhen the app runs and then crashes.
- Emergency RepairDown now? Front of the queue.
- Website RepairThe whole bench.
- 403 Forbidden403 Forbidden
- Redirect LoopToo many redirects
- Memory ExhaustedMemory exhausted
Find out which server deserves the finger.
Send the symptom, get a look within one business day and a flat quote from the developer who's fixed this exact thing more times than either of us can count.