feat(build): raise BuildKit cache-mount budget for remote builders (#112)
Some checks failed
ci/crow/cron/weekly-audit-missing/3 Pipeline was successful
ci/crow/cron/weekly-audit-missing/4 Pipeline was successful
ci/crow/cron/weekly-audit-missing/5 Pipeline was successful
ci/crow/cron/weekly-audit-missing/6 Pipeline was successful
ci/crow/cron/weekly-audit-missing/7 Pipeline was successful
ci/crow/cron/weekly-audit-missing/8 Pipeline was successful
ci/crow/cron/weekly-audit-missing/9 Pipeline was successful
ci/crow/cron/weekly-audit-missing/10 Pipeline was successful
ci/crow/cron/weekly-audit-missing/11 Pipeline was successful
ci/crow/cron/weekly-audit-missing/12 Pipeline was successful
ci/crow/cron/weekly-audit-missing/13 Pipeline was successful
ci/crow/cron/weekly-audit-missing/14 Pipeline was successful
ci/crow/cron/weekly-audit-missing/15 Pipeline was successful
ci/crow/cron/weekly-audit-missing/16 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/3 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/4 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/1 Pipeline was canceled
ci/crow/cron/weekly-rebuild-missing/6 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/7 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/8 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/2 Pipeline was canceled
ci/crow/cron/weekly-rebuild-missing/5 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/10 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/11 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/12 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/13 Pipeline was successful
ci/crow/cron/weekly-rebuild-missing/9 Pipeline was canceled
ci/crow/cron/weekly-rebuild-missing/14 Pipeline was successful
ci/crow/cron/process-updates/15 Pipeline failed
ci/crow/cron/process-updates/11 Pipeline was canceled
ci/crow/cron/process-updates/16 Pipeline was canceled
ci/crow/cron/process-updates/12 Pipeline was successful
ci/crow/cron/process-updates/5 Pipeline was successful
ci/crow/cron/process-updates/6 Pipeline was successful
ci/crow/cron/process-updates/18 Pipeline was successful
ci/crow/cron/process-updates/17 Pipeline was canceled
ci/crow/cron/process-updates/1 Pipeline was canceled
ci/crow/cron/process-updates/2 Pipeline was canceled
ci/crow/cron/process-updates/8 Pipeline was canceled
ci/crow/manual/build-all-versions-install-deps/2 Pipeline was successful
ci/crow/cron/process-updates/9 Pipeline was canceled
ci/crow/cron/process-updates/3 Pipeline was canceled
ci/crow/cron/process-updates/7 Pipeline was successful
ci/crow/cron/process-updates/10 Pipeline was canceled
ci/crow/manual/build-all-versions/7 Pipeline was canceled
ci/crow/manual/build-all-versions/6 Pipeline failed
ci/crow/manual/build-all-versions/8 Pipeline was canceled
ci/crow/manual/build-all-versions/5 Pipeline was canceled
ci/crow/cron/process-updates/4 Pipeline was canceled
ci/crow/cron/process-updates/14 Pipeline was canceled
ci/crow/cron/process-updates/13 Pipeline was successful

BuildKit's default GC caps the ephemeral cache tier — `RUN --mount=type=cache` mounts, local build context, git checkouts — at a hardcoded **512 MB** (shown as `488.3 MiB` in `buildx inspect`).
Across our 7-distro build matrix that fills instantly and forces re-downloads of system + R packages on every rebuild.

## Changes

- **`docker/buildkitd.toml`** (new) — GC config passed via `--config` when creating the `docker-container` builders `artemis` (amd64) and `gaia` (arm64):
  - cache-mount tier: **512 MB → 8 GB**, retained 7 days
  - total cache bounded at **40 GB** with **20 GB min-free**
- **`justfile`** — document the `--config docker/buildkitd.toml` flag on the builder-create commands so a recreate does not silently revert to the 512 MB default.

## Notes

- Limits are absolute (not `%`) because the two hosts differ ~6× in free space (Hetzner ~42 GiB free vs Mac mini ~279 GiB). The 20 GB min-free is the safety valve on the disk-tight Hetzner host; the old default wanted 64 GiB free, which does not exist there.
- Both live builders were already recreated with this config and verified running.

Reviewed-on: #112
This commit is contained in:
Patrick Schratz 2026-07-08 08:22:03 +00:00 committed by Patrick Schratz
commit 09536cddd6
2 changed files with 45 additions and 2 deletions

41
docker/buildkitd.toml Normal file
View file

@ -0,0 +1,41 @@
# BuildKit GC config for the remote buildx builders (artemis/amd64, gaia/arm64).
#
# Apply when creating the docker-container builders:
# docker buildx create --name artemis --driver docker-container \
# --config docker/buildkitd.toml ssh://<user@host-amd64>
# docker buildx create --name gaia --driver docker-container \
# --config docker/buildkitd.toml ssh://<user@host-arm64>
#
# Why: BuildKit's default GC caps the ephemeral cache tier — RUN
# --mount=type=cache mounts, local build context, git checkouts — at a
# hardcoded 512 MB (shown as "488.3 MiB" in `buildx inspect`). Across our
# 7-distro build matrix that fills instantly and forces re-downloads of
# system + R packages every rebuild. The first rule below raises that tier.
#
# Limits are absolute (not %) on purpose: artemis and gaia have very
# different free space (Hetzner ~42 GiB free vs Mac mini ~279 GiB), so a
# percentage would mean wildly different real budgets. minFreeSpace = 20 GB
# keeps the tight Hetzner host safe while staying modest on the Mac mini.
[worker.oci]
gc = true
# Tier 1 — ephemeral caches (cache mounts, local context, git checkouts).
# Raised from the 512 MB default to 8 GB, retained for 7 days so weekly
# rebuilds reuse downloaded packages instead of re-fetching them.
[[worker.oci.gcpolicy]]
filters = [
"type==source.local",
"type==exec.cachemount",
"type==source.git.checkout",
]
keepDuration = "168h"
maxUsedSpace = "8GB"
# Tier 2 — everything else (image layers, RUN exec results). Bounds the
# whole buildkit cache and always leaves 20 GB free on the host disk.
[[worker.oci.gcpolicy]]
all = true
reservedSpace = "2GB"
maxUsedSpace = "40GB"
minFreeSpace = "20GB"

View file

@ -10,8 +10,10 @@
# - buildx builders named `artemis` (amd64) and `gaia` (arm64), created with the
# docker-container driver (runs BuildKit on the remote host's docker daemon over
# SSH). The default `remote` driver does NOT work with an ssh:// docker host.
# docker buildx create --name artemis --driver docker-container ssh://<user@host-amd64>
# docker buildx create --name gaia --driver docker-container ssh://<user@host-arm64>
# Pass --config docker/buildkitd.toml so BuildKit's GC keeps a usable cache
# (the default caps the cache-mount tier at 512 MB, forcing re-downloads).
# docker buildx create --name artemis --driver docker-container --config docker/buildkitd.toml ssh://<user@host-amd64>
# docker buildx create --name gaia --driver docker-container --config docker/buildkitd.toml ssh://<user@host-arm64>
# - exported secrets: B2_S3_ACCESS_KEY, B2_S3_SECRET_KEY, PGPASS (GITHUB_PAT optional)
#
# Overridable (env or `just VAR=… rebuild …`):