Resource footprint: dated full-stack receipts
Canonical dated docker stats for Epure plus PostgreSQL. Other pages link here. Do not average the rows.
On this pageShowHide
This page is the canonical home for dated resource footprint receipts on epure.sh. Other essays and guides link here instead of pasting MiB rows again. You are here for what the two-container install used in RAM. The numbers below are first-party docker stats for the Epure container and PostgreSQL together. They are dated named-box receipts, not a vendor minimum and not a promise for your VPS.
Use this page if
- You need the dated full-stack rows before you quote a MiB number.
- You want the method next to the result: host, image, both containers, idle versus a short ingest burst.
- You are editing a blog, /vs page, or listicle and need one citation URL for Epure RAM.
Do not use this page if
- You need a guaranteed RAM ceiling for a production SLO. We are not publishing one.
- You want the full narrative argument against Sentry's self-host floor. That essay is Sentry self-hosted RAM.
- You need the five-tool install matrix. That lives on Self-hosted research.
Answer first
Epure idle footprint on dated named boxes is small relative to Sentry's published ~16 GB self-host floor. Quote the receipt that matches your box class. Sentry documents COMPOSE_PROFILES=errors-only; disclose that official lighter profile before you differentiate, then compare fleet size honestly. Errors-only still keeps a large compose. It is not two containers.
Method
- Command:
docker statson classic Compose once the stack is healthy and idle. Both containers. App-only RSS is not a row. - VPS row: 2026-09-23, 2 vCPU / 769 MiB Linux (Alibaba Cloud), image
ghcr.io/epure-sh/epure:latestplus Postgres. - Desktop row: 2026-09-12 Docker Desktop idle, combined app + Postgres. That row does not split the two containers.
- Load row: a short store-ingest burst on the same VPS, about 280-330 requests/s. It is not a sustained load test.
- Peer floors cited for context only: Sentry ~16 GB RAM + 16 GB swap (full suite docs); GlitchTip 256 MB min / 512 MB recommended (their install docs). Re-check those vendor pages before you buy hardware.
Results
Idle
- When
- 2026-09-23
- Host
- 2 vCPU / 769 MiB Linux VPS, Alibaba Cloud
- RSS
- Epure ~5 MiB, Postgres ~48 MiB, ~53 MiB combined
Short store-ingest burst
- When
- 2026-09-23, same VPS
- Host
- About 280-330 requests/s, not a sustained load test
- RSS
- Epure under ~12 MiB; Postgres stayed the heavier process
Idle
- When
- 2026-09-12
- Host
- Docker Desktop
- RSS
- ~82 MiB combined, app + Postgres
| Dimension | When | Host | RSS |
|---|---|---|---|
| Idle | 2026-09-23 | 2 vCPU / 769 MiB Linux VPS, Alibaba Cloud | Epure ~5 MiB, Postgres ~48 MiB, ~53 MiB combined |
| Short store-ingest burst | 2026-09-23, same VPS | About 280-330 requests/s, not a sustained load test | Epure under ~12 MiB; Postgres stayed the heavier process |
| Idle | 2026-09-12 | Docker Desktop | ~82 MiB combined, app + Postgres |
VPS image: ghcr.io/epure-sh/epure:latest with Postgres on classic Compose. Desktop is a laptop Docker VM. Do not blend ~53 MiB and ~82 MiB into one floor. Neither row is a vendor minimum. Re-measure on your box.
Sentry context
When a comparison mentions Sentry RAM, disclose COMPOSE_PROFILES=errors-only first. That profile drops traces, replay, and profiling services and keeps issue-related services. It remains a large fleet. Epure's dated rows above are two containers (epure + postgres:16-alpine) with no Redis or Kafka in the OSS compose. The fair write-up of the suite floor lives on Sentry self-hosted RAM. This page owns the Epure table.
Limits
- Docker Desktop numbers are not a VPS. A VPS number is not every VPS.
- Idle is not a crash loop. Disk growth from retention is a different bill. Backups and upgrades.
- Spike valve and hard ingest caps are different stories. Spike valve.
- No Redis or Kafka in this compose, so do not compare a two-process sample to a 20-container suite as if the jobs matched. Scope: Envelope.
- Cloud list prices on this site are published; Cloud is not live as of 2026-10-05. Self-host is what these receipts measure.
How to remeasure
docker stats --no-streamafter healthy idle, or host RSS for theepureandpostgresprocesses.- Record box SKU, image tags (
EPURE_IMAGE), date, and whether the UI was open. - Undated screenshots are not site facts. Do not publish them as floors.
Prerequisites to re-run
- Compose from Installation.
GET /healthreturns{"status":"ok"}. - A project DSN and one SDK. JS CI pin is
@sentry/browseror@sentry/node7.120.0.
Verification
Confirm the install before you compare RSS. curl /health for ok. Send one exception. Expect HTTP 202 and one Issues row. Then docker stats on both containers after healthy idle. Verify.
Troubleshooting
If /health fails or the issue never appears, that is an install or DSN problem, not a footprint result. Invalid DSN, Events not showing.
Takeaway
One table, dated rows, link everywhere else. Remeasure when images change. Disclose Sentry errors-only when the comparison mentions Sentry RAM. Narrative essays that argue the floor: Under 512 MB, Two-container, Sentry self-hosted RAM.