build-cran-binaries/.crow/prune-stale-flat-binaries.yaml
pat-s 5c4cca6cce fix: verify the metadata database certificate with sslmode=verify-full (#235)
Part of https://codefloe.com/rpkgs/bincraft/issues/225. Stacked on #234.

Every connection to the metadata database used `sslmode = 'require'`, which encrypts but does not authenticate the server, so anyone able to impersonate `r-binaries.devxy.io:15432` would receive `PGPASS`.

## Changes

- `.crow/build-all-versions.yaml`, `process-updates.yaml`, `prune-stale-flat-binaries.yaml`, `reindex.yaml`, `weekly-rebuild-reindex.yaml`: `metadata_db_sslmode = 'verify-full'` in every bincraft call.
- `local/*.R`: the same for bincraft calls; scripts that call `DBI::dbConnect()` directly also pass `sslrootcert = bincraft::postgres_sslrootcert("verify-full")`, because libpq looks for `~/.postgresql/root.crt` rather than the system trust store.
- `README.md`: note that the endpoint serves a Let's Encrypt certificate and how to verify it.
- air reformatted three `=` assignments to `<-` in `local/missing-cran-packages-db.R` and `local/packages-without-any-binary.R` while the hook ran on those files.

## Ordering: merge last

1. devxy/k8s-talos#417 must be live: today the server presents an in-cluster certificate from a private CA, which `verify-full` rejects.
2. https://codefloe.com/rpkgs/bincraft/pulls/226 must be released: `local/install-bincraft.R` installs the latest tag, and older releases neither pass a root certificate nor export `postgres_sslrootcert()`.
3. #234 must be merged.

Merging early fails every database connection in CI, and since #234 per-minor reindex failures fail the pipeline.

Reviewed-on: #235
2026-10-17 07:45:46 +00:00

251 lines
8.4 KiB
YAML

# Remove flat-slot binaries recorded unable to load under some R minor, then
# re-index.
#
# The list comes from `abi_flat_safety`, so it names the objects the binaries
# themselves report as broken, not the ones a source heuristic suspects. That
# distinction is the whole cost of the operation: the heuristic calls 93% of
# compiled packages risky and would delete roughly 3000 objects per slot to fix
# about 230.
#
# `upload_single_binary()` has written a non-portable package only to the
# per-minor slot since v5.3.0, so this is a one-time cleanup of what was
# uploaded before that. The flat slot is what every client that cannot be
# routed by R minor receives, and such a binary was built under whichever R was
# primary at the time, so it dies at load under any minor that dropped a symbol
# it uses:
#
# base64enc.so: undefined symbol: SETLENGTH
#
# Deleting the flat object does not create a hole. The source fallback
# republishes the CRAN source under the same key, so the package stays
# installable everywhere and compiles against the client's own R, while routed
# clients keep getting the per-minor binary.
#
# Dry run unless PRUNE_APPLY is 'yes'. Start with an explicit package list:
#
# crow pipeline create devxy/build-cran-binaries \
# --var prune_flat=ubuntu-2604-amd64 \
# --var PRUNE_PACKAGES="base64enc lobstr" \
# --var PRUNE_APPLY=yes
variables:
# A manual pipeline creation instantiates every file in .crow/, so the default
# must match no matrix row.
prune_flat:
description: "Prune target: a specific <os>-<arch>, 'all' for every slot, or 'none'."
options:
- none
- all
- alpine-323-amd64
- alpine-323-arm64
- alpine-324-amd64
- alpine-324-arm64
- redhat-8-amd64
- redhat-8-arm64
- redhat-9-amd64
- redhat-9-arm64
- redhat-10-amd64
- redhat-10-arm64
- ubuntu-2204-amd64
- ubuntu-2204-arm64
- ubuntu-2404-amd64
- ubuntu-2404-arm64
- ubuntu-2604-amd64
- ubuntu-2604-arm64
default: none
PRUNE_PACKAGES:
description: 'Space-separated package names. Empty means every package the classifier cache calls risky.'
default: ''
PRUNE_APPLY:
description: "Set to 'yes' to actually delete. Anything else is a dry run."
options:
- 'no'
- 'yes'
default: 'no'
when:
- event: manual
evaluate: 'prune_flat == "all" || prune_flat == "${OS}-${ARCH}"'
skip_clone: true
labels:
group: rpkgs-${ARCH}
matrix:
include:
- OS: alpine-323
ARCH: amd64
CODENAME: alpine323
IMG: alpine:3.23
- OS: alpine-323
ARCH: arm64
CODENAME: alpine323
IMG: alpine:3.23
- OS: alpine-324
ARCH: amd64
CODENAME: alpine324
IMG: alpine:3.24
- OS: alpine-324
ARCH: arm64
CODENAME: alpine324
IMG: alpine:3.24
- OS: redhat-8
ARCH: amd64
CODENAME: rhel8
IMG: redhat:8
- OS: redhat-8
ARCH: arm64
CODENAME: rhel8
IMG: redhat:8
- OS: redhat-9
ARCH: amd64
CODENAME: rhel9
IMG: redhat:9
- OS: redhat-9
ARCH: arm64
CODENAME: rhel9
IMG: redhat:9
- OS: redhat-10
ARCH: amd64
CODENAME: rhel10
IMG: redhat:10
- OS: redhat-10
ARCH: arm64
CODENAME: rhel10
IMG: redhat:10
- OS: ubuntu-2204
ARCH: amd64
CODENAME: jammy
IMG: ubuntu:jammy
- OS: ubuntu-2204
ARCH: arm64
CODENAME: jammy
IMG: ubuntu:jammy
- OS: ubuntu-2404
ARCH: amd64
CODENAME: noble
IMG: ubuntu:noble
- OS: ubuntu-2404
ARCH: arm64
CODENAME: noble
IMG: ubuntu:noble
- OS: ubuntu-2604
ARCH: amd64
CODENAME: resolute
IMG: ubuntu:resolute
- OS: ubuntu-2604
ARCH: arm64
CODENAME: resolute
IMG: ubuntu:resolute
steps:
- name: 'Prune stale flat binaries'
image: reg.devxy.io/rpkgs/build-env-${IMG}
pull: true
environment:
OTEL_R_TRACES_EXPORTER: none
OTEL_R_LOGS_EXPORTER: none
OTEL_R_METRICS_EXPORTER: none
B2_S3_ACCESS_KEY:
from_secret: B2_S3_ACCESS_KEY
B2_S3_SECRET_KEY:
from_secret: B2_S3_SECRET_KEY
PGPASS:
from_secret: PGPASS
REPO_RO_TOKEN:
from_secret: REPO_RO_TOKEN
GIT_USER: pat-s
R_LIBS_USER: /tmp/R-pkgs
# Pipeline variables are substituted into the YAML, not exported, so they
# have to be bound here to be readable as env vars in the script below.
PRUNE_ARCH: ${ARCH}
PRUNE_CODENAME: ${CODENAME}
PRUNE_PACKAGES: ${PRUNE_PACKAGES}
PRUNE_APPLY: ${PRUNE_APPLY}
commands:
- git clone -q https://pat-s:$$REPO_RO_TOKEN@git.devxy.io/devxy/build-cran-binaries.git .
- mkdir -p /tmp/R-pkgs
- /opt/R/4.5.3/bin/Rscript local/install-bincraft.R
- |
set -eu
APPLY=""
if [ "$$PRUNE_APPLY" = "yes" ]; then
APPLY="--apply"
fi
/opt/R/4.5.3/bin/Rscript local/prune-stale-flat-binaries.R \
"$$PRUNE_ARCH" "$$PRUNE_CODENAME" $$APPLY $$PRUNE_PACKAGES
backend_options:
docker:
resources:
requests:
memory: 3Gi
cpu: 1000m
# Deleting objects without re-indexing is worse than leaving them: the
# published index keeps naming files that no longer exist, so every client
# gets a 404 instead of a working source fallback. The re-index is what
# republishes them as CRAN source, so it belongs in the same run.
- name: 'Re-index the slot'
image: reg.devxy.io/rpkgs/build-env-${IMG}
pull: true
environment:
OTEL_R_TRACES_EXPORTER: none
OTEL_R_LOGS_EXPORTER: none
OTEL_R_METRICS_EXPORTER: none
B2_S3_ACCESS_KEY:
from_secret: B2_S3_ACCESS_KEY
B2_S3_SECRET_KEY:
from_secret: B2_S3_SECRET_KEY
PGPASS:
from_secret: PGPASS
R_LIBS_USER: /tmp/R-pkgs
PRUNE_APPLY: ${PRUNE_APPLY}
commands:
- |
set -eu
if [ "$$PRUNE_APPLY" != "yes" ]; then
echo "Dry run: nothing was deleted, so there is nothing to re-index."
exit 0
fi
/opt/R/4.5.3/bin/R -q -e 'library(bincraft); upload_package_index(s3_endpoint = "https://s3.eu-central-003.backblazeb2.com", s3_region = "eu-central-003", s3_bucket = "devxy-rpkgs-binaries", s3_access_key_id = Sys.getenv("B2_S3_ACCESS_KEY"), s3_secret_access_key = Sys.getenv("B2_S3_SECRET_KEY"))'
FAILED_MINORS=""
for RBIN in /opt/R/[0-9]*/bin/R; do
RMINOR=$(basename "$(dirname "$(dirname "$RBIN")")" | cut -d. -f1-2)
/opt/R/4.5.3/bin/R -q -e "library(bincraft); upload_package_index(r_minor = '$RMINOR', s3_endpoint = 'https://s3.eu-central-003.backblazeb2.com', s3_region = 'eu-central-003', s3_bucket = 'devxy-rpkgs-binaries', s3_access_key_id = Sys.getenv('B2_S3_ACCESS_KEY'), s3_secret_access_key = Sys.getenv('B2_S3_SECRET_KEY'), metadata_db_host = 'r-binaries.devxy.io', metadata_db_name = 'build_metadata', metadata_db_port = 15432L, metadata_db_user = 'rpkgs', metadata_db_password = Sys.getenv('PGPASS'), metadata_db_sslmode = 'verify-full')" || FAILED_MINORS="$FAILED_MINORS $RMINOR"
done
if [ -n "$FAILED_MINORS" ]; then
echo "Could not republish the per-minor index for:$FAILED_MINORS" >&2
exit 1
fi
backend_options:
docker:
resources:
requests:
memory: 3Gi
cpu: 1000m
# The edge caches indexes for ~370 days, so without this the deletion is
# invisible to clients and they keep being handed the removed binary.
- name: 'Purge CDN cache'
image: reg.devxy.io/docker.io/library/alpine:3.24
environment:
BUNNYNET_API_KEY:
from_secret: BUNNYNET_API_KEY
# Same zones as reindex.yaml: the production pull zone id, the alliance
# hostname and the staging zone, which shares the B2 origin.
BUNNY_PULLZONES: '3857050 cran.allianceswisspass.devxy.io cran-rpkgs-test.b-cdn.net'
PRUNE_APPLY: ${PRUNE_APPLY}
commands:
- |
set -eu
if [ "$$PRUNE_APPLY" != "yes" ]; then
echo "Dry run: nothing changed at the origin, so no purge is needed."
exit 0
fi
apk add --no-cache -q bash curl jq
bash scripts/purge_cdn_zone.sh "$BUNNYNET_API_KEY" $BUNNY_PULLZONES
# Run it even when a per-minor re-index above failed: the generic index
# and the objects were still replaced, and a stale edge keeps them hidden.
when:
- status: [success, failure]