Error: read ECONNRESET

at TCP.onStreamRead (node:internal/stream_base_commons:216:20) {

errno: -104,

code: 'ECONNRESET',

syscall: 'read'

}

Server has closed the connection.

This happened also yesterday at the exact same time

I'm losing my patience here...this is the third day in a row and i have heard NOTHING from the team. PLEASE INVESTIGATE ASAP

Apologies for the delay on this. To pin down whether this is the same EU West TCP issue as your May 6 thread or something new, could you share the exact UTC timestamps of the ECONNRESET errors over the past 3 days (a few samples per day is enough) and confirm which service is hitting them? With those we can correlate against EU West proxy and database telemetry.

May 18, 2026 — 10 errors (window: 07:39–13:21 UTC)

| # | Timestamp (UTC) | Type |

| --- | --- | --- |

| 1 | 2026-05-18T07:39:38Z | Redis ECONNRESET |

| 2 | 2026-05-18T07:40:39Z | Redis ECONNRESET |

| 3 | 2026-05-18T07:40:54Z | Redis ECONNRESET |

| 4 | 2026-05-18T07:45:29Z | Redis ECONNRESET |

| 5 | 2026-05-18T07:45:33Z | Redis ECONNRESET |

| 6 | 2026-05-18T07:46:42Z | Redis ECONNRESET |

| 7 | 2026-05-18T07:47:43Z | Axios ECONNRESET (webhook POST) |

| 8 | 2026-05-18T08:25:33Z | Redis ECONNRESET |

| 9 | 2026-05-18T08:26:18Z | Redis ECONNRESET |

| 10 | 2026-05-18T13:21:52Z | Axios ECONNRESET (webhook POST) |

May 19, 2026 — 12 errors (window: 07:56–08:06 UTC)

| # | Timestamp (UTC) | Type |

| --- | --- | --- |

| 1 | 2026-05-19T07:56:49Z | Redis ECONNRESET |

| 2 | 2026-05-19T08:00:05Z | Redis ECONNRESET |

| 3 | 2026-05-19T08:00:21Z | Redis ECONNRESET |

| 4 | 2026-05-19T08:01:36Z | Redis ECONNRESET |

| 5 | 2026-05-19T08:01:51Z | Redis ECONNRESET |

| 6 | 2026-05-19T08:02:06Z | Redis ECONNRESET |

| 7 | 2026-05-19T08:02:37Z | Redis ECONNRESET |

| 8 | 2026-05-19T08:02:52Z | Redis ECONNRESET |

| 9 | 2026-05-19T08:03:07Z | Redis ECONNRESET |

| 10 | 2026-05-19T08:04:07Z | Redis ECONNRESET |

| 11 | 2026-05-19T08:04:53Z | Redis ECONNRESET |

| 12 | 2026-05-19T08:06:54Z | Redis ECONNRESET |

May 20, 2026 — 34 errors (window: 08:08–12:19 UTC)

| # | Timestamp (UTC) | Type |

| --- | --- | --- |

| 1 | 2026-05-20T08:08:12Z | Redis ECONNRESET |

| 2 | 2026-05-20T08:08:27Z | Redis ECONNRESET |

| 3 | 2026-05-20T08:09:12Z | Redis ECONNRESET |

| 4 | 2026-05-20T08:09:58Z | Redis ECONNRESET |

| 5 | 2026-05-20T08:11:44Z | Redis ECONNRESET |

| 6 | 2026-05-20T08:11:59Z | Redis ECONNRESET |

| 7 | 2026-05-20T08:13:14Z | Redis ECONNRESET |

| 8 | 2026-05-20T08:13:29Z | Redis ECONNRESET |

| 9 | 2026-05-20T08:14:11Z | Axios ECONNRESET (webhook POST) |

| 10–15 | 08:17:01Z – 08:20:33Z | Redis ECONNRESET (6×) |

| 16–22 | 09:39:11Z – 09:46:04Z | Redis ECONNRESET (7×, second burst) |

| 23–24 | 11:25:46Z – 11:26:01Z | Redis ECONNRESET |

| 25 | 2026-05-20T11:28:41Z | Axios ECONNRESET (webhook POST) |

| 26 | 2026-05-20T11:55:47Z | Redis ECONNRESET |

| 27–34 | 12:19:00Z – 12:19:52Z | Redis ECONNRESET (8×, rapid burst ~7s apart) |

Hey, I just spent some time digging in here.

Sorry for the run of issues on EU West, both the recent three mornings and going back to March.

What we're seeing:

May 20 lines up with platform recovery from the May 19/20 GCP account outage (post-mortem: https://blog.railway.com/p/incident-report-may-19-2026-gcp-account-outage). Services on GCP whose private-network route caches went cold during the outage hit re-connection failures while routes were being re-resolved. That accounts for the 08:08 to 12:19 bursts.

May 18 and 19 mornings tie into a separate known issue on our private network in EU West that we're actively tracking. The 07:56 to 08:06 UTC window on May 19 specifically lines up with some alerts on our end. The May 6 fix doesn't apply here since you're on *.railway.internal, not the public TCP proxy.

We have active escalations open around these issues and we're prioritizing improving routing performance in EU West that caused the May 18 and 19 issues via increased capacity and resilience.

Thanks Mykal. When can I expect an update about the ongoing issues? Are you going to publish a post-mortem regarding private networking in EU-West as well?

We don't have a specific timeline for the next update on the EU West private networking improvements yet, but this is actively being worked on. Regarding a post-mortem, we publish those for major platform-wide incidents like the GCP account outage I linked. We can't commit to a separate post-mortem for the private networking issue at this time, but if there are further developments we'll update you here.