an hour ago
Hello Railway team,
The disk usage graph for our Redis volume has grown from approximately 70 MB to 360 MB, without showing a corresponding decrease.
However, a read-only inspection on November 17, 2026, showed that:
- The filesystem reports only 1.3 MB used.
- The stored files occupy approximately 92 KiB.
- The only data file is a Redis snapshot of approximately 86 KiB.
Could you please clarify:
- Why does the dashboard report significantly more usage than the filesystem?
- Does the metric include underlying block allocation or space that has not been reclaimed?
- Is this reported usage also used for billing?
- Is any supported configuration or maintenance needed to reclaim this space and prevent continued growth?
We would like to understand this discrepancy before changing Redis persistence settings or performing any destructive maintenance.
Thank you.
1 Reply
an hour ago
The dashboard figure is not your files. It's the number of storage blocks allocated to the volume at the storage layer underneath your filesystem, so it will always read higher than df or du inside the container. Space freed inside the container only goes back to that layer once it's discarded, and volumes allocate in 16 KB blocks while the filesystem frees 4 KB blocks. Partly freed blocks can't be returned, so workloads that keep rewriting small files, like Redis snapshot saves, can hold allocation well above the data you can see. A volume that has been resized can also grow for a while because the filesystem sets up metadata for the new capacity, and a redeploy stops that growth. Your snapshot and filesystem are fine, and the figure won't drop to match df.
Billing does use this metric. Volume usage is billed on the same allocated figure, averaged over the billing period, at $0.15 per GB-month. Your Redis volume currently reads about 377 MB, which comes to roughly $0.06 a month.
You don't need to change Redis persistence settings or do any destructive maintenance. Running fstrim inside the container will fail with "Operation not permitted", and that's expected.
Status changed to Awaiting User Response Railway • about 1 hour ago