From a4b274f28148d7d257007c6abd1d51e6cbd69227 Mon Sep 17 00:00:00 2001 From: pat-s Date: Mon, 13 Jul 2026 09:31:29 +0000 Subject: [PATCH] fix(ci): bound pkgcache _metadata growth to stop macmini disk-fill (#110) ## Problem The arm64 `build-all` pipeline fills the macmini (gaia) host disk despite an 8h prune. Root cause is not images or job volumes: it is the persistent dep-cache volume, specifically `pkgcache/R/pkgcache/_metadata`, which grew to ~165 GB. `{pkgcache}` mints a new content hash for the "patched" binaries repo on every PACKAGES change, so each per-package build writes a fresh ~70 MB `pkgs-.rds` (+ `patched-/`) that is never evicted (2407 snapshots observed). When the disk hits 100% OrbStack stops and the on-host prune can no longer connect to the daemon, so it never self-heals. ## Change (Workstream A of the disk-fill fix) - Add `trim_pkgcache_metadata()` to `local/r-minor-helpers.R`: keeps the newest `keep` (default 20) `patched-*`/`pkgs-*.rds` entries under `_metadata`, deleting only entries older than `min_age_secs` (default 600s) so it never races the up-to-4 concurrent split-jobs sharing the volume. Preserves `pkg/` downloads and the stable CRAN/BioC/INLA repo dirs. No-op when `R_PKG_CACHE_DIR` is empty (amd64) or `_metadata` is absent (first run). - Call it every 25 packages inside the build loop in `local/build-all.R`. - Add a defensive start-of-run cleanup of `_metadata/patched-*` + `pkgs-*.rds` to the two workflows that mount the persistent volume (`build-all-versions.yaml`, `build-all-versions-install-deps.yaml`). Only these paths are touched; `process-updates.yaml`/`weekly-rebuild-missing.yaml` (no persistent volume) are unchanged. Follow-ups (separate workstreams): on-host self-healing prune watcher + OrbStack disk cap (ansible), and Prometheus/Grafana alerting (k8s-talos). Upstream: bincraft patched-repo hash churn is the true source fix. New unit tests (6) for the helper; full suite 24/24 green. Reviewed-on: https://git.devxy.io/devxy/build-cran-binaries/pulls/110 --- .crow/build-all-versions-install-deps.yaml | 8 ++- .crow/build-all-versions.yaml | 9 ++- local/build-all.R | 10 ++++ local/r-minor-helpers.R | 39 +++++++++++++ local/tests/test-trim-pkgcache.R | 68 ++++++++++++++++++++++ 5 files changed, 130 insertions(+), 4 deletions(-) create mode 100644 local/tests/test-trim-pkgcache.R diff --git a/.crow/build-all-versions-install-deps.yaml b/.crow/build-all-versions-install-deps.yaml index 8649ca0..e0cca86 100644 --- a/.crow/build-all-versions-install-deps.yaml +++ b/.crow/build-all-versions-install-deps.yaml @@ -4,7 +4,7 @@ # they are merged with build-all-versions' identical declarations. variables: target_arch: - description: "Architecture to build." + description: 'Architecture to build.' options: - amd64 - arm64 @@ -27,7 +27,7 @@ variables: - "noble" default: "3.24" R_VERSION: - description: "Primary R version under /opt/R." + description: 'Primary R version under /opt/R.' options: - 4.5.3 - 4.4.3 @@ -74,6 +74,10 @@ steps: commands: # one-time full wipe to fix corrupted .so files from previous failed builds # - rm -rf /mnt/cache/R-pkgs + # Clear churny pkgcache metadata left by a prior crashed run (the "patched" + # repo mints a new hash per PACKAGES change -> unbounded pkgs-*.rds/patched-*). + # Keep pkg/ downloads and the stable CRAN/BioC/INLA repo dirs. + - rm -rf /mnt/cache/pkgcache/R/pkgcache/_metadata/patched-* /mnt/cache/pkgcache/R/pkgcache/_metadata/pkgs-*.rds || true - mkdir -p /mnt/cache/pkgcache /mnt/cache/R-pkgs /mnt/cache/ccache /mnt/cache/packages - git clone -q https://pat-s:$$REPO_RO_TOKEN@git.devxy.io/devxy/build-cran-binaries.git . # Pin the same bincraft version the build steps use, so the precomputed diff --git a/.crow/build-all-versions.yaml b/.crow/build-all-versions.yaml index 615d273..3932811 100644 --- a/.crow/build-all-versions.yaml +++ b/.crow/build-all-versions.yaml @@ -5,7 +5,7 @@ # Skip list lives in local/excluded-packages.json (read by local/build-all.R). variables: target_arch: - description: "Architecture to build." + description: 'Architecture to build.' options: - amd64 - arm64 @@ -31,7 +31,7 @@ variables: - "resolute" default: "3.24" R_VERSION: - description: "Primary R version under /opt/R." + description: 'Primary R version under /opt/R.' options: - 4.5.3 - 4.4.3 @@ -116,6 +116,11 @@ steps: - ${ARCH}-binaries-r-dep-cache-${OS}-${OS_VERSION//./}:/mnt/cache commands: - git clone -q https://pat-s:$$REPO_RO_TOKEN@git.devxy.io/devxy/build-cran-binaries.git . + # Clear churny pkgcache metadata left by a prior crashed run (the "patched" + # repo mints a new hash per PACKAGES change -> unbounded pkgs-*.rds/patched-*). + # Keep pkg/ downloads and the stable CRAN/BioC/INLA repo dirs. Within-run + # growth is bounded separately by trim_pkgcache_metadata() in build-all.R. + - rm -rf /mnt/cache/pkgcache/R/pkgcache/_metadata/patched-* /mnt/cache/pkgcache/R/pkgcache/_metadata/pkgs-*.rds || true - mkdir -p /mnt/cache/pkgcache /mnt/cache/R-pkgs /mnt/cache/ccache /mnt/cache/packages # The primary pass must not rely on build-all-versions-install-deps having # run on *this* agent: depends_on only orders the steps, but the cache diff --git a/local/build-all.R b/local/build-all.R index 7983be3..1df1f83 100644 --- a/local/build-all.R +++ b/local/build-all.R @@ -122,6 +122,10 @@ s3_cache <- readRDS("/mnt/cache/packages/s3_cache.rds") sprintf("S3 cache: %s files", length(s3_cache)) n <- nrow(chunk) +# Every `trim_every` packages, bound the pkgcache _metadata dir so a full-platform +# run does not accumulate thousands of ~70 MB snapshots and fill the host disk. +# No-op on amd64 (R_PKG_CACHE_DIR is empty / cache not persisted). +trim_every <- 25L mapply( function(pkg, ver, sens, i) { cat(sprintf("[%d/%d] %s_%s (r_minor_sensitive=%s)\n", i, n, pkg, ver, sens)) @@ -147,6 +151,12 @@ mapply( upload = TRUE, store_build_metadata = TRUE ) + if (i %% trim_every == 0L) { + removed <- trim_pkgcache_metadata() + if (removed > 0L) { + cat(sprintf(" [pkgcache trim] removed %d stale _metadata entries\n", removed)) + } + } }, chunk$Package, chunk$Version, diff --git a/local/r-minor-helpers.R b/local/r-minor-helpers.R index 65aafdf..6f01c29 100644 --- a/local/r-minor-helpers.R +++ b/local/r-minor-helpers.R @@ -31,3 +31,42 @@ parse_build_args <- function(args) { ncpus = as.integer(pos[3L]) ) } + +# Bound the {pkgcache} metadata dir, which otherwise grows without limit: the +# "patched" repo mints a new content hash on every PACKAGES change, so each build +# writes a fresh ~70 MB _metadata/pkgs-.rds (+ patched-/) that is +# never reused. Keep the `keep` newest entries by mtime; only remove entries +# older than `min_age_secs`, so a concurrent split-job's in-flight files are +# never deleted (each build uses a unique hash, so aged entries are +# unreferenced). Stable repo dirs (CRAN-*, BioC*, INLA-*) and pkg/ downloads are +# not matched and thus preserved. Returns the number of entries removed. +trim_pkgcache_metadata <- function(cache_dir = Sys.getenv("R_PKG_CACHE_DIR"), + keep = 20L, + min_age_secs = 600) { + meta <- file.path(cache_dir, "R", "pkgcache", "_metadata") + if (!nzchar(cache_dir) || !dir.exists(meta)) { + return(0L) + } + entries <- c( + Sys.glob(file.path(meta, "patched-*")), + Sys.glob(file.path(meta, "pkgs-*.rds")) + ) + if (length(entries) == 0L) { + return(0L) + } + info <- file.info(entries) + order_new_first <- order(info$mtime, decreasing = TRUE) + ranked <- entries[order_new_first] + ranked_mtime <- info$mtime[order_new_first] + if (length(ranked) <= keep) { + return(0L) + } + candidates <- ranked[(keep + 1L):length(ranked)] + candidate_age <- as.numeric(Sys.time()) - as.numeric(ranked_mtime[(keep + 1L):length(ranked)]) + removable <- candidates[candidate_age >= min_age_secs] + if (length(removable) == 0L) { + return(0L) + } + unlink(removable, recursive = TRUE, force = TRUE) + length(removable) +} diff --git a/local/tests/test-trim-pkgcache.R b/local/tests/test-trim-pkgcache.R new file mode 100644 index 0000000..1a5b418 --- /dev/null +++ b/local/tests/test-trim-pkgcache.R @@ -0,0 +1,68 @@ +source(file.path("..", "r-minor-helpers.R")) + +# Build a fake _metadata dir under a temp R_PKG_CACHE_DIR. Each entry's mtime is +# set to `age_secs` in the past so we can exercise the age gate deterministically. +make_meta <- function(patched = 0L, pkgs = 0L, keep_repos = TRUE, age_secs = 3600) { + root <- tempfile("pkgcache-") + meta <- file.path(root, "R", "pkgcache", "_metadata") + dir.create(meta, recursive = TRUE) + old <- Sys.time() - age_secs + mk_dir <- function(p) { dir.create(p); Sys.setFileTime(p, old); p } + mk_file <- function(p) { writeLines("x", p); Sys.setFileTime(p, old); p } + for (i in seq_len(patched)) mk_dir(file.path(meta, sprintf("patched-%03d", i))) + for (i in seq_len(pkgs)) mk_file(file.path(meta, sprintf("pkgs-%03d.rds", i))) + if (keep_repos) { + mk_dir(file.path(meta, "CRAN-075c426938")) + mk_dir(file.path(meta, "BioCsoft-1ac964ed6c")) + mk_file(file.path(meta, "bioc-sysreqs.dcf.gz")) + mk_dir(file.path(root, "R", "pkgcache", "pkg")) # downloads, must survive + } + root +} + +n_churn <- function(root) { + meta <- file.path(root, "R", "pkgcache", "_metadata") + length(Sys.glob(file.path(meta, "patched-*"))) + + length(Sys.glob(file.path(meta, "pkgs-*.rds"))) +} + +test_that("empty cache_dir is a no-op", { + expect_identical(trim_pkgcache_metadata("", keep = 5L, min_age_secs = 0), 0L) +}) + +test_that("missing _metadata dir is a no-op", { + expect_identical( + trim_pkgcache_metadata(tempfile("absent-"), keep = 5L, min_age_secs = 0), + 0L + ) +}) + +test_that("fewer than keep entries removes nothing", { + root <- make_meta(patched = 2L, pkgs = 2L) + expect_identical(trim_pkgcache_metadata(root, keep = 20L, min_age_secs = 0), 0L) + expect_identical(n_churn(root), 4L) +}) + +test_that("trims down to keep newest, leaving churn == keep", { + root <- make_meta(patched = 30L, pkgs = 30L) # 60 churn entries, all old + removed <- trim_pkgcache_metadata(root, keep = 20L, min_age_secs = 0) + expect_identical(removed, 40L) + expect_identical(n_churn(root), 20L) +}) + +test_that("entries younger than min_age_secs are protected", { + root <- make_meta(patched = 30L, pkgs = 0L, keep_repos = FALSE, age_secs = 60) + # keep=5 would drop 25, but all are 60s old < 600s gate -> nothing removed + expect_identical(trim_pkgcache_metadata(root, keep = 5L, min_age_secs = 600), 0L) + expect_identical(n_churn(root), 30L) +}) + +test_that("stable repo dirs and pkg downloads are never touched", { + root <- make_meta(patched = 30L, pkgs = 30L) + trim_pkgcache_metadata(root, keep = 0L, min_age_secs = 0) + meta <- file.path(root, "R", "pkgcache", "_metadata") + expect_true(dir.exists(file.path(meta, "CRAN-075c426938"))) + expect_true(dir.exists(file.path(meta, "BioCsoft-1ac964ed6c"))) + expect_true(file.exists(file.path(meta, "bioc-sysreqs.dcf.gz"))) + expect_true(dir.exists(file.path(root, "R", "pkgcache", "pkg"))) +})