Skip to content
YapHub Discord login blocked by HTTP 429 on Railway US East; users cannot access bot
diese-tech
OP

a month ago

YapHub is a Discord bot that creates temporary voice rooms when users join configured lobbies. We expected it to remain available, but Discord API requests began failing and the replacement deployment crashed during login, leaving users unable to use the bot. At our last check on October 12 at 00:34:40 EDT, the patched process was running but had not logged into Discord. Please investigate the outbound connection/IP restriction and advise whether Railway can help restore access and prevent recurrence.

Resources and timing

Runtime logs show 25 failed room-creation attempts across two guilds from October 11 at 19:11:55 through 21:14:47 EDT. Creation and cleanup resumed at 21:24 EDT without a deployment; the last observed successful creation was 21:50:15 EDT. After redeployment on October 12, login failed repeatedly around 00:26 EDT.

Evidence and attempts

The crashed deployment's Railway runtime logs show the failure during initial login, before room creation:

2026-10-12 04:26:18,632 INFO discord.client logging in using static token
File "/app/.venv/lib/python3.13/site-packages/discord/http.py", line 844, in static_login
    data = await self.request(Route('GET', '/users/@me'))
discord.errors.HTTPException: 429 Too Many Requests (error code: 0): You are being blocked from accessing our API temporarily due to exceeding global rate limits. Refer to https://discord.com/developers/docs/topics/rate-limits for more information.

The patched deployment logged this at 04:32:03 UTC on October 12:

2026-10-12 04:32:03,040 WARNING yaphub startup login_rate_limited attempt=1 retry_after=2688.0 pause_source=retry_after_header scope=None global=None via=None cf_ray=a49360e2fd174cad-IAD content_type=application/json

At the last log check, October 12 at 04:34:40 UTC / 00:34:40 EDT, no successful login was observed. The process was waiting for the supplied interval; its next attempt was expected around 05:16:51 UTC / 01:16:51 EDT. Subsequent recovery is unverified.

Checks and changes actually performed:

  • DNS queries for discord.com at the first failures returned NOERROR and IPv4 addresses. A short network-flow sample recorded outbound HTTPS and inbound traffic with no recorded drop cause. These are connection observations, not complete request accounting.
  • PR #30 added room-creation cooldowns. Login remained unprotected.
  • PR #32 added login backoff, respecting Retry-After and closing/resetting failed HTTP resources. The deployment stopped the crash loop but was still waiting for Discord access.
  • Local Python 3.11 tests (399) and lint passed; CI passed. Production uses Python 3.13.15. These do not establish API availability.

Code: https://github.com/diese-tech/lab-yaphub/pull/30 and https://github.com/diese-tech/lab-yaphub/pull/32

No token rotation, region migration, outbound-IP change or paid-plan upgrade was attempted.

Why contact support and uncertainty

The AI assistant first recommended support to identify the historical public egress IP and unconfirmed cause. It recommended urgent escalation after the long Retry-After response prevented users accessing the bot.

An edge/shared-egress-IP restriction is a hypothesis, supported by the response metadata and recovery without code changes. We have not proved that another Railway customer caused it. The logs say “global rate limits”; missing scope headers and incomplete HTTP accounting prevent definitive attribution. DNS and returning network traffic weaken a simple transport-outage explanation. Public status pages showed no matching incident, which does not exclude a localized restriction.

Please correlate the identified deployments and UTC windows with their public egress assignment, check for known Discord restrictions affecting that outbound path, and advise on supported recovery and prevention options. Preserve the existing volume and data; this is a diagnostic request, not authorization to delete, migrate or modify resources.

Prepared with Railway's support template (v1).

$10 Bounty

5 Replies

Railway
BOT

a month ago

This thread has been opened as a bounty so the community can help solve it.

Status changed to Open Railway • about 1 month ago


diese-tech
OP

a month ago

AI-assisted; this update is based on observed Railway production logs.

Thanks for reviewing the patch. The existing Retry-After wait completed successfully on October 12: logs show startup login_recovered at 05:16:51 UTC, a Gateway connection at 05:16:51 UTC, and Logged in as at 05:16:55 UTC. This was deployment c6a03ae6-beaf-475f-8e98-94b803b508f0. Successful post-recovery room creation and cleanup have not yet been verified.

Agreed on persisting the cooldown across restarts; we plan to include that in the recovery hardening tracked in https://github.com/diese-tech/lab-yaphub/issues/31. It is not implemented yet. To correct the earlier AI explanation: the 429 rejection and retry interval are confirmed, but the underlying restriction mechanism and any shared-IP interference remain unconfirmed. We also understand that static outbound IPs are not guaranteed dedicated.

Could Railway staff correlate the reported failure window and CF-Ray a49360e2fd174cad-IAD with this service's historical public egress IP and any known outbound restrictions? Login and Gateway connectivity recovered, but the cause remains unresolved. This is a request for diagnosis; please preserve the existing volume and data.


a month ago

railway IPs are shared, if you got rate limited it probably just means someone on the same block as you has been. you can try enabling outbound static IPs, but they are still shared, just static and shared amongst less services


diese-tech
OP

a month ago

AI-assisted

Thanks, milo. That helps. I understand the static IP option is still shared. I haven't changed the region, enabled static IPs, or rotated the bot token.

YapHub started having the same issue again today. It handles the main VC flow for SAL, YSL, and a few other servers, so this takes down the feature people are using the bot for. I want to understand what's causing the recurring block and how to keep it stable.

The latest login failures are from deployment 9f0936b3-49ba-4417-a3a4-4cc1f96b75ae. These timestamps are UTC:

2026-10-12 22:54:59,402: Retry-After 487 seconds; CF-Ray a499b0851d0c9c6c-IAD

2026-10-12 23:03:06,599: local fallback 120 seconds; CF-Ray a499bc6a38154a55-IAD

2026-10-12 23:05:06,722: Retry-After 3495 seconds; CF-Ray a499bf58fbe3d6d9-IAD

The 120 seconds came from the bot's fallback, not Discord. The latest Discord wait ends at 2026-10-13T00:03:21.722Z, or 8:03:21 PM EDT tonight.

I've deployed PR #33 as afe98f65-13c8-40da-a4fd-801341a204ad. It now pauses all affected REST requests, including cleanup, and saves the cooldown across restarts. I had Codex verify the SQLite backup and preserve that deadline before deployment. There's one replica with no deployment overlap, and the process check found one bot worker. The local Windows check didn't find another matching bot process, although that doesn't rule out other machines or unknown deployments.

The process is alive, but the bot hasn't logged back in yet. /healthz returned 200, /readyz returned 503, and status showed the preserved deadline with zero outgoing API attempts since this deployment started. Automatic recovery for users waiting in the lobby is the next release under issue #31.

Could you check the historical egress assignments for these timestamps and CF-Ray IDs? Your explanation makes shared-egress interference plausible, but I'd like to confirm whether that's what happened here. Is there a supported dedicated outbound IPv4 option beyond the shared static-IP feature?

Please keep the volume and data intact. I'm asking for diagnosis and options, not a migration or resource changes.


a month ago

i can’t do that; I’m just a moderator- when i used to run discord bots I very rarely ran into this issue but usually waiting an hour or two then retrying fixed it. static IPs is ur best bet


diese-tech
OP

a month ago

AI-assisted

Thanks, milo—I appreciate you sharing your experience.

YapHub is back online. Discord login recovered after the wait, and I fixed a separate startup bug in the bot. I verified the complete flow in two Discord servers: joining a lobby creates a voice room, moves the user, posts the controls, and deletes the room after everyone leaves.

The bot now saves cooldowns across restarts and can resume queued work after an interruption. I haven’t enabled static outbound IPs or changed regions.

The original cause of the recurring 429 blocks is still unconfirmed. For others running Discord bots on Railway, have you experienced similar recurring blocks, and did enabling static outbound IPs help?


Welcome!

Sign in to your Railway account to join the conversation.

Loading...