a day ago
We're experiencing what appears to be degraded database performance and were hoping you could help investigate.
Around the time the issue started, it looks like the database cache may have been unexpectedly cleared or reset. The memory usage suddenly dropped significantly (from around 30 GB to ~15 GB) without any intentional restart or configuration change on our side.
We also did not deploy any application or database changes during this period. There were no deployments, configuration updates, or maintenance performed on our side that could explain this behavior.
Since then we've observed:
- Noticeably degraded database performance
- Increased response times (especially p95/p99)
- A spike in request errors
- Memory usage dropping abruptly, which makes us suspect the database cache was lost or invalidated
We've attached screenshots of the relevant metrics.
Attachments
4 Replies
a day ago
We’ve been experiencing a similar issue, but the Railway team doesn’t really seem to acknowledge it xd. Everything seems fine right now, though. It feels like something broke in the background and then somehow fixed itself.
Since our own cgroup limits are not the constraint, this looks like host-level (global) memory pressure from other workloads on the same host reclaiming our page cache.
Could you please:
- Confirm whether the host this service runs on is under memory pressure / oversubscribed, and what changed around today 11:00am KST.
- Migrate this service to a less contended host.
This is production and actively impacting users.. really need help!
a day ago
I am writing to inquire about my account status. My services have been suspended/blocked, but according to my dashboard panel, I still have a Credits Remaining balance of $0.92. I that somethig linked to this thread too or the problem?
I noticed that Days Remaining shows 0 days, and the system message states the trial expires when either condition is met. However, I wanted to clarify if there is any way to utilize the remaining credit balance before the account forces an upgrade, or if the system automatically cuts off access the exact moment the day counter hits zero regardless of funds.
I have attached Screenshot 2026-07-05 203404.png showing my current trial status dropdown for reference.
Could you please clarify this policy and let me know the best way to resume my services?
Thank you,
Our database is currently stuck during startup, and the deploy logs repeatedly show only:
Mounting volume on: /var/lib/containers/railwayapp/bind-mounts/c857e751-f589-4666-a7af-9bb4fec7115b/vol_9pj2uu9r8hdzpp0q
This message repeats continuously, and PostgreSQL never starts.
Before this happened, PostgreSQL entered recovery mode and repeatedly logged:
FATAL: the database system is not yet accepting connections
DETAIL: Consistent recovery state has not been reached.