In the Railway dashboard, click the Observability tab at the top instead of the per-service log panel. Set the time range to cover when your alert emails were sent, and use a search term from those alerts (e.g., a unique substring in the log message) so you know you’re matching the right events.

Filter by environment and service

In Log Explorer, filter to your environment first, then select only your main service so you’re not mixing logs from Redis and the Tailscale forwarder.

If you still don’t see entries, briefly clear service filters and check whether those same messages show up tagged as coming from another process, which would indicate mislabeling in the UI rather than true loss.

Use railway logs from the CLI

Install or open the Railway CLI and run something like: railway logs --service --since 24h.

Cross‑check timestamps and message content against your email alerts; if they appear in CLI but not the web viewer, you’re dealing with a viewer issue, not missing storage.

If logs truly aren’t stored

If the logs don’t appear in the CLI at all:

Confirm log level and logger config in your main service (e.g., ensure you didn’t accidentally move those critical alerts to a different logger or level when adding Redis/Tailscale). This kind of misconfiguration would cause alerts to fire but structured logs to go elsewhere or be dropped.

Check any new environment variables or log-related settings you added for Redis or Tailscale forwarder so you’re not redirecting stdout/stderr away from Railway’s log capture (for instance, wrapping your app in another process that doesn’t forward logs properly).

I need help from the Railway team on this. The deployment info for the service that has these logs is:

Project: AAF Organizational Pipeline

Environment: production

Service: aaf-tool

Deployment ID: 17a74a1b-0893-4cb8-adee-55c79221b8f0

I have email that says that at or just before 6:07:25am PT there was a log entry containing "Warning: timeout before entering data". There would be other log entries just before and after that entry containing related info. But when I use the CLI to get those logs, I see no entries with timestamps in the entire 6:07 minute:

$ railway logs 17a74a1b-0893-4cb8-adee-55c79221b8f0 --since 2026-07-29T13:06:00Z --until 2026-07-29T13:09:00Z --json

{"message":"Processing home page request","level":"error","timestamp":"2026-07-29T13:06:15.221636585Z"}

{"timestamp":"2026-07-29T13:06:15.932873744Z","message":"Processing job page request for job:EX2bNX4BHX (org:RvvRk4JAXX)","level":"error"}

{"message":"Processing rerun page request for job:EX2bNX4BHX (org:RvvRk4JAXX)","timestamp":"2026-07-29T13:06:18.031847450Z","level":"error"}

{"timestamp":"2026-07-29T13:06:18.031853710Z","level":"error","message":"Rerunning job:EX2bNX4BHX"}

{"message":"Submitted from front end: job:yt4BjFtQpd (org:yt4BjFtQpd): 20 cases","level":"error","timestamp":"2026-07-29T13:06:18.039376638Z"}

{"message":"Reloaded proxy for job:EX2bNX4BHX: Completed, 20/20 cases in 56.581331948s","timestamp":"2026-07-29T13:06:18.069440808Z","level":"error"}

{"message":"Processing job page request for job:yt4BjFtQpd (org:RvvRk4JAXX)","timestamp":"2026-07-29T13:06:18.175028513Z","level":"error"}

{"level":"error","message":"Grabbed job:yt4BjFtQpd from the waiting queue and put it in the processing queue","timestamp":"2026-07-29T13:06:21.497597780Z"}

{"timestamp":"2026-07-29T13:06:21.497603760Z","level":"error","message":"There are 0 jobs left in the waiting queue"}

{"message":"Starting processing of job:yt4BjFtQpd: Submitted, 0/20 cases in 0s","timestamp":"2026-07-29T13:06:21.500182775Z","level":"error"}

{"timestamp":"2026-07-29T13:06:21.500188515Z","message":"Reloaded proxy for job:yt4BjFtQpd: Processing, 0/20 cases in 0s","level":"error"}

{"message":"Starting web session Web test session for job:yt4BjFtQpd with driver timeout 8s","timestamp":"2026-07-29T13:06:22.547684231Z","level":"error"}

{"level":"error","timestamp":"2026-07-29T13:06:22.547689461Z","message":"Starting check session Web test session for job:yt4BjFtQpd separating checks by 1.75s"}

{"level":"error","message":"Processing job page request for job:yt4BjFtQpd (org:RvvRk4JAXX)","timestamp":"2026-07-29T13:08:03.539321222Z"}

What I do see at that time, when I look in the log viewer, is a bunch of messages from the Redis service. I need the Railway team to determine whether this is an log egress bandwidth problem and advise me how I can avoid this happening in future.

Are Redis/Tailscale running as separate services, or is Tailscale wrapped into your main service's start command/entrypoint?

Separate services.