mirror of
https://github.com/Xahau/xahaud.git
synced 2025-12-06 17:27:52 +00:00
refactor: remove OverlayFS delta caching entirely
THE GREAT CULLING: Remove all OverlayFS and delta caching logic. After extensive investigation and testing, we determined that OverlayFS file-level layering is fundamentally incompatible with ccache's access patterns: - ccache opens files with O_RDWR → kernel must provide writable file handle - OverlayFS must copy files to upper layer immediately (can't wait) - Even with metacopy=on, metadata-only files still appear in upper layer - Result: ~366MB deltas instead of tiny incremental diffs The fundamental constraint: cannot have all three of: 1. Read-only lower layer (for base sharing) 2. Writable file handles (for O_RDWR) 3. Minimal deltas (for efficient caching) Changes: - Removed all OverlayFS mounting/unmounting logic - Removed workspace and registry tracking - Removed delta creation and restoration - Removed use-deltas parameter - Simplified to direct tar/extract workflow Before: 726 lines across cache actions After: 321 lines (-55% reduction) Benefits: - ✅ Simpler architecture (direct tar/extract) - ✅ More maintainable (less code, less complexity) - ✅ More reliable (fewer moving parts) - ✅ Same performance (base-only was already used) - ✅ Clear path forward (restic/borg for future optimization) Current state works great: - Build times: 20-30 min → 2-5 min (80% improvement) - Cache sizes: ~323-609 MB per branch (with zst compression) - S3 costs: acceptable for current volume If bandwidth costs become problematic, migrate to restic/borg for chunk-level deduplication (completely different architecture).
This commit is contained in:
Nicholas Dudfield
368
.github/actions/xahau-actions-cache-save/action.yml
vendored
368
.github/actions/xahau-actions-cache-save/action.yml
vendored
@@ -1,5 +1,5 @@
|
||||
name: 'Xahau Cache Save (S3 + OverlayFS)'
|
||||
description: 'Drop-in replacement for actions/cache/save using S3 and OverlayFS for delta caching'
|
||||
name: 'Xahau Cache Save (S3)'
|
||||
description: 'Drop-in replacement for actions/cache/save using S3 storage'
|
||||
|
||||
inputs:
|
||||
path:
|
||||
@@ -16,10 +16,6 @@ inputs:
|
||||
description: 'S3 region'
|
||||
required: false
|
||||
default: 'us-east-1'
|
||||
use-deltas:
|
||||
description: 'Enable delta caching (download/upload incremental changes). Set to false for base-only caching.'
|
||||
required: false
|
||||
default: 'true'
|
||||
# Note: Composite actions can't access secrets.* directly - must be passed from workflow
|
||||
aws-access-key-id:
|
||||
description: 'AWS Access Key ID for S3 access'
|
||||
@@ -31,7 +27,7 @@ inputs:
|
||||
runs:
|
||||
using: 'composite'
|
||||
steps:
|
||||
- name: Save cache to S3 with OverlayFS delta
|
||||
- name: Save cache to S3
|
||||
shell: bash
|
||||
env:
|
||||
AWS_ACCESS_KEY_ID: ${{ inputs.aws-access-key-id }}
|
||||
@@ -40,12 +36,11 @@ runs:
|
||||
S3_REGION: ${{ inputs.s3-region }}
|
||||
CACHE_KEY: ${{ inputs.key }}
|
||||
TARGET_PATH: ${{ inputs.path }}
|
||||
USE_DELTAS: ${{ inputs.use-deltas }}
|
||||
run: |
|
||||
set -euo pipefail
|
||||
|
||||
echo "=========================================="
|
||||
echo "Xahau Cache Save (S3 + OverlayFS)"
|
||||
echo "Xahau Cache Save (S3)"
|
||||
echo "=========================================="
|
||||
echo "Target path: ${TARGET_PATH}"
|
||||
echo "Cache key: ${CACHE_KEY}"
|
||||
@@ -53,346 +48,63 @@ runs:
|
||||
echo ""
|
||||
|
||||
# Normalize target path (expand tilde and resolve to absolute path)
|
||||
# This ensures consistent path comparison with the mount registry
|
||||
if [[ "${TARGET_PATH}" == ~* ]]; then
|
||||
# Expand tilde manually (works even if directory doesn't exist yet)
|
||||
TARGET_PATH="${HOME}${TARGET_PATH:1}"
|
||||
fi
|
||||
echo "Normalized target path: ${TARGET_PATH}"
|
||||
echo ""
|
||||
|
||||
# Find the cache workspace from mount registry
|
||||
MOUNT_REGISTRY="/tmp/xahau-cache-mounts.txt"
|
||||
|
||||
if [ ! -f "${MOUNT_REGISTRY}" ]; then
|
||||
echo "⚠️ No cache mounts found (mount registry doesn't exist)"
|
||||
echo "This usually means cache restore was not called, or there was no cache to restore."
|
||||
# Check if target directory exists
|
||||
if [ ! -d "${TARGET_PATH}" ]; then
|
||||
echo "⚠️ Target directory does not exist: ${TARGET_PATH}"
|
||||
echo "Skipping cache save."
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Find entry for this path
|
||||
# Format: path:workspace:matched_key:primary_key:exact_match:use_deltas
|
||||
# Bootstrap mode: path:bootstrap:key:key:false:true/false (workspace="bootstrap")
|
||||
CACHE_WORKSPACE=""
|
||||
MATCHED_KEY=""
|
||||
PRIMARY_KEY=""
|
||||
EXACT_MATCH=""
|
||||
REGISTRY_USE_DELTAS=""
|
||||
# Use static base name (one base per key, immutable)
|
||||
S3_BASE_KEY="s3://${S3_BUCKET}/${CACHE_KEY}-base.tar.zst"
|
||||
|
||||
while IFS=: read -r mount_path mount_workspace mount_matched_key mount_primary_key mount_exact_match mount_use_deltas; do
|
||||
if [ "${mount_path}" = "${TARGET_PATH}" ]; then
|
||||
CACHE_WORKSPACE="${mount_workspace}"
|
||||
MATCHED_KEY="${mount_matched_key}"
|
||||
PRIMARY_KEY="${mount_primary_key}"
|
||||
EXACT_MATCH="${mount_exact_match}"
|
||||
REGISTRY_USE_DELTAS="${mount_use_deltas}"
|
||||
break
|
||||
fi
|
||||
done < "${MOUNT_REGISTRY}"
|
||||
|
||||
if [ -z "${CACHE_WORKSPACE}" ] && [ -z "${MATCHED_KEY}" ]; then
|
||||
echo "⚠️ No cache entry found for path: ${TARGET_PATH}"
|
||||
echo "This usually means cache restore was not called for this path."
|
||||
echo "Skipping cache save."
|
||||
# Check if base already exists (immutability - first write wins)
|
||||
if aws s3 ls "${S3_BASE_KEY}" --region "${S3_REGION}" >/dev/null 2>&1; then
|
||||
echo "⚠️ Cache already exists: ${S3_BASE_KEY}"
|
||||
echo "Skipping upload (immutability - first write wins, like GitHub Actions)"
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Cache save completed (already exists)"
|
||||
echo "=========================================="
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Determine cache mode
|
||||
if [ "${CACHE_WORKSPACE}" = "bootstrap" ]; then
|
||||
CACHE_MODE="bootstrap"
|
||||
PRIMARY_KEY="${MATCHED_KEY}" # In bootstrap, matched_key field contains primary key
|
||||
echo "Cache mode: BOOTSTRAP (first build for this key)"
|
||||
echo "Primary key: ${PRIMARY_KEY}"
|
||||
elif [ "${EXACT_MATCH}" = "false" ]; then
|
||||
CACHE_MODE="partial-match"
|
||||
echo "Cache mode: PARTIAL MATCH (restore-key used)"
|
||||
echo "Cache workspace: ${CACHE_WORKSPACE}"
|
||||
echo "Matched key from restore: ${MATCHED_KEY}"
|
||||
echo "Primary key (will save new base): ${PRIMARY_KEY}"
|
||||
else
|
||||
CACHE_MODE="exact-match"
|
||||
echo "Cache mode: EXACT MATCH (cache hit)"
|
||||
echo "Cache workspace: ${CACHE_WORKSPACE}"
|
||||
echo "Matched key: ${MATCHED_KEY}"
|
||||
fi
|
||||
echo "Use deltas: ${REGISTRY_USE_DELTAS}"
|
||||
# Create tarball
|
||||
BASE_TARBALL=$(mktemp /tmp/xahau-cache-XXXXXX.tar.zst)
|
||||
|
||||
echo "Creating cache tarball..."
|
||||
tar -cf - -C "${TARGET_PATH}" . | zstd -3 -T0 -q -o "${BASE_TARBALL}"
|
||||
|
||||
BASE_SIZE=$(du -h "${BASE_TARBALL}" | cut -f1)
|
||||
echo "✓ Cache tarball created: ${BASE_SIZE}"
|
||||
echo ""
|
||||
|
||||
# Handle different cache modes
|
||||
if [ "${CACHE_MODE}" = "bootstrap" ]; then
|
||||
# Bootstrap: Save entire cache as base layer (no OverlayFS was used)
|
||||
echo "Bootstrap mode: Creating initial base layer from ${TARGET_PATH}"
|
||||
# Upload to S3
|
||||
echo "Uploading cache to S3..."
|
||||
echo " Key: ${CACHE_KEY}-base.tar.zst"
|
||||
|
||||
BASE_TARBALL="/tmp/xahau-cache-base-$$.tar.zst"
|
||||
echo "Creating base tarball..."
|
||||
tar -cf - -C "${TARGET_PATH}" . | zstd -3 -T0 -q -o "${BASE_TARBALL}"
|
||||
aws s3api put-object \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--key "${CACHE_KEY}-base.tar.zst" \
|
||||
--body "${BASE_TARBALL}" \
|
||||
--tagging 'type=base' \
|
||||
--region "${S3_REGION}" \
|
||||
>/dev/null
|
||||
|
||||
BASE_SIZE=$(du -h "${BASE_TARBALL}" | cut -f1)
|
||||
echo "✓ Base tarball created: ${BASE_SIZE}"
|
||||
echo ""
|
||||
echo "✓ Uploaded: ${S3_BASE_KEY}"
|
||||
|
||||
# Use static base name (one base per key, immutable)
|
||||
S3_BASE_KEY="s3://${S3_BUCKET}/${PRIMARY_KEY}-base.tar.zst"
|
||||
|
||||
# Check if base already exists (immutability - first write wins)
|
||||
if aws s3 ls "${S3_BASE_KEY}" --region "${S3_REGION}" >/dev/null 2>&1; then
|
||||
echo "⚠️ Base layer already exists: ${S3_BASE_KEY}"
|
||||
echo "Skipping upload (immutability - first write wins, like GitHub Actions)"
|
||||
else
|
||||
echo "Uploading base layer to S3..."
|
||||
echo " Key: ${PRIMARY_KEY}-base.tar.zst"
|
||||
|
||||
aws s3api put-object \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--key "${PRIMARY_KEY}-base.tar.zst" \
|
||||
--body "${BASE_TARBALL}" \
|
||||
--tagging 'type=base' \
|
||||
--region "${S3_REGION}" \
|
||||
>/dev/null
|
||||
|
||||
echo "✓ Uploaded: ${S3_BASE_KEY}"
|
||||
fi
|
||||
|
||||
# Cleanup
|
||||
rm -f "${BASE_TARBALL}"
|
||||
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Bootstrap cache save completed"
|
||||
echo "Base size: ${BASE_SIZE}"
|
||||
echo "Cache key: ${PRIMARY_KEY}"
|
||||
echo "=========================================="
|
||||
exit 0
|
||||
|
||||
elif [ "${CACHE_MODE}" = "partial-match" ]; then
|
||||
# Partial match: Save merged view as new base ONLY (no delta)
|
||||
# The delta is relative to the OLD base, not the NEW base we're creating
|
||||
echo "Partial match mode: Saving new base layer for primary key"
|
||||
echo "Note: Delta will NOT be saved (it's relative to old base)"
|
||||
|
||||
BASE_TARBALL="/tmp/xahau-cache-base-$$.tar.zst"
|
||||
echo "Creating base tarball from merged view..."
|
||||
tar -cf - -C "${CACHE_WORKSPACE}/merged" . | zstd -3 -T0 -q -o "${BASE_TARBALL}"
|
||||
|
||||
BASE_SIZE=$(du -h "${BASE_TARBALL}" | cut -f1)
|
||||
echo "✓ Base tarball created: ${BASE_SIZE}"
|
||||
echo ""
|
||||
|
||||
# Use static base name (one base per key, immutable)
|
||||
S3_BASE_KEY="s3://${S3_BUCKET}/${PRIMARY_KEY}-base.tar.zst"
|
||||
|
||||
# Check if base already exists (immutability - first write wins)
|
||||
if aws s3 ls "${S3_BASE_KEY}" --region "${S3_REGION}" >/dev/null 2>&1; then
|
||||
echo "⚠️ Base layer already exists: ${S3_BASE_KEY}"
|
||||
echo "Skipping upload (immutability - first write wins, like GitHub Actions)"
|
||||
else
|
||||
echo "Uploading new base layer to S3..."
|
||||
echo " Key: ${PRIMARY_KEY}-base.tar.zst"
|
||||
|
||||
aws s3api put-object \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--key "${PRIMARY_KEY}-base.tar.zst" \
|
||||
--body "${BASE_TARBALL}" \
|
||||
--tagging 'type=base' \
|
||||
--region "${S3_REGION}" \
|
||||
>/dev/null
|
||||
|
||||
echo "✓ Uploaded: ${S3_BASE_KEY}"
|
||||
fi
|
||||
|
||||
# Cleanup
|
||||
rm -f "${BASE_TARBALL}"
|
||||
|
||||
# Unmount and cleanup
|
||||
echo ""
|
||||
echo "Cleaning up..."
|
||||
if mount | grep -q "${CACHE_WORKSPACE}/merged"; then
|
||||
sudo umount "${CACHE_WORKSPACE}/merged" || {
|
||||
echo "⚠️ Warning: Failed to unmount ${CACHE_WORKSPACE}/merged"
|
||||
echo "Attempting lazy unmount..."
|
||||
sudo umount -l "${CACHE_WORKSPACE}/merged" || true
|
||||
}
|
||||
fi
|
||||
rm -rf "${CACHE_WORKSPACE}"
|
||||
|
||||
# Remove from registry
|
||||
if [ -f "${MOUNT_REGISTRY}" ]; then
|
||||
grep -v "^${TARGET_PATH}:" "${MOUNT_REGISTRY}" > "${MOUNT_REGISTRY}.tmp" 2>/dev/null || true
|
||||
mv "${MOUNT_REGISTRY}.tmp" "${MOUNT_REGISTRY}" 2>/dev/null || true
|
||||
fi
|
||||
|
||||
echo "✓ Cleanup completed"
|
||||
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Partial match cache save completed"
|
||||
echo "New base created for: ${PRIMARY_KEY}"
|
||||
echo "Base size: ${BASE_SIZE}"
|
||||
if [ "${REGISTRY_USE_DELTAS}" = "true" ]; then
|
||||
echo "Next exact-match build will create deltas from this base"
|
||||
else
|
||||
echo "Next exact-match build will reuse this base (base-only mode)"
|
||||
fi
|
||||
echo "=========================================="
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# For exact-match ONLY: Save delta (if use-deltas enabled)
|
||||
if [ "${CACHE_MODE}" = "exact-match" ]; then
|
||||
# If deltas are disabled, just cleanup and exit
|
||||
if [ "${REGISTRY_USE_DELTAS}" != "true" ]; then
|
||||
echo "ℹ️ Delta caching disabled (use-deltas: false)"
|
||||
echo "Base already exists for this key, nothing to save."
|
||||
|
||||
# Unmount and cleanup
|
||||
echo ""
|
||||
echo "Cleaning up..."
|
||||
if mount | grep -q "${CACHE_WORKSPACE}/merged"; then
|
||||
sudo umount "${CACHE_WORKSPACE}/merged" 2>/dev/null || true
|
||||
fi
|
||||
rm -rf "${CACHE_WORKSPACE}"
|
||||
|
||||
# Remove from registry
|
||||
if [ -f "${MOUNT_REGISTRY}" ]; then
|
||||
grep -v "^${TARGET_PATH}:" "${MOUNT_REGISTRY}" > "${MOUNT_REGISTRY}.tmp" 2>/dev/null || true
|
||||
mv "${MOUNT_REGISTRY}.tmp" "${MOUNT_REGISTRY}" 2>/dev/null || true
|
||||
fi
|
||||
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Cache save completed (base-only mode)"
|
||||
echo "=========================================="
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Check if upper layer has any changes
|
||||
if [ -z "$(ls -A ${CACHE_WORKSPACE}/upper 2>/dev/null)" ]; then
|
||||
echo "ℹ️ No changes detected in upper layer (cache is unchanged)"
|
||||
echo "Skipping delta upload to save bandwidth."
|
||||
|
||||
# Still unmount and cleanup
|
||||
echo ""
|
||||
echo "Cleaning up..."
|
||||
sudo umount "${CACHE_WORKSPACE}/merged" 2>/dev/null || true
|
||||
rm -rf "${CACHE_WORKSPACE}"
|
||||
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Cache save completed (no changes)"
|
||||
echo "=========================================="
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Show delta statistics
|
||||
echo "Delta layer statistics:"
|
||||
echo " Files changed: $(find ${CACHE_WORKSPACE}/upper -type f 2>/dev/null | wc -l)"
|
||||
echo " Delta size: $(du -sh ${CACHE_WORKSPACE}/upper 2>/dev/null | cut -f1)"
|
||||
echo ""
|
||||
|
||||
# Create delta tarball from upper layer
|
||||
echo "Creating delta tarball..."
|
||||
DELTA_TARBALL="/tmp/xahau-cache-delta-$$.tar.zst"
|
||||
|
||||
tar -cf - -C "${CACHE_WORKSPACE}/upper" . | zstd -3 -T0 -q -o "${DELTA_TARBALL}"
|
||||
|
||||
DELTA_SIZE=$(du -h "${DELTA_TARBALL}" | cut -f1)
|
||||
echo "✓ Delta tarball created: ${DELTA_SIZE}"
|
||||
echo ""
|
||||
|
||||
# Upload timestamped delta (no overwrites = zero concurrency issues)
|
||||
TIMESTAMP=$(date +%Y%m%d%H%M%S)
|
||||
COMMIT_SHA=$(git rev-parse --short HEAD 2>/dev/null || echo "unknown")
|
||||
|
||||
# Use PRIMARY_KEY for delta (ensures deltas match their base)
|
||||
S3_DELTA_TIMESTAMPED="s3://${S3_BUCKET}/${PRIMARY_KEY}-delta-${TIMESTAMP}-${COMMIT_SHA}.tar.zst"
|
||||
|
||||
echo "Uploading timestamped delta to S3..."
|
||||
echo " Key: ${PRIMARY_KEY}-delta-${TIMESTAMP}-${COMMIT_SHA}.tar.zst"
|
||||
|
||||
# Upload with tag (deltas cleaned up inline - keep last 1)
|
||||
aws s3api put-object \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--key "${PRIMARY_KEY}-delta-${TIMESTAMP}-${COMMIT_SHA}.tar.zst" \
|
||||
--body "${DELTA_TARBALL}" \
|
||||
--tagging 'type=delta-archive' \
|
||||
--region "${S3_REGION}" \
|
||||
>/dev/null
|
||||
|
||||
echo "✓ Uploaded: ${S3_DELTA_TIMESTAMPED}"
|
||||
|
||||
# Inline cleanup: Keep only latest delta (the one we just uploaded)
|
||||
echo ""
|
||||
echo "Cleaning up old deltas (keeping only latest)..."
|
||||
|
||||
# List all deltas for this key, sorted by LastModified (oldest first)
|
||||
ALL_DELTAS=$(aws s3api list-objects-v2 \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--prefix "${PRIMARY_KEY}-delta-" \
|
||||
--region "${S3_REGION}" \
|
||||
--query 'sort_by(Contents, &LastModified)[*].Key' \
|
||||
--output json 2>/dev/null || echo "[]")
|
||||
|
||||
DELTA_COUNT=$(echo "${ALL_DELTAS}" | jq 'length' 2>/dev/null || echo "0")
|
||||
|
||||
if [ "${DELTA_COUNT}" -gt 1 ]; then
|
||||
# Keep last 1 (newest), delete all older ones (all except last 1 = [0:-1])
|
||||
OLD_DELTAS=$(echo "${ALL_DELTAS}" | jq -r '.[0:-1][]' 2>/dev/null)
|
||||
|
||||
if [ -n "${OLD_DELTAS}" ]; then
|
||||
DELETE_COUNT=$((DELTA_COUNT - 1))
|
||||
echo " Found ${DELETE_COUNT} old delta(s) to delete"
|
||||
|
||||
# Create delete batch request JSON
|
||||
DELETE_OBJECTS=$(echo "${OLD_DELTAS}" | jq -R -s -c 'split("\n") | map(select(length > 0)) | map({Key: .}) | {Objects: ., Quiet: true}' 2>/dev/null)
|
||||
|
||||
if [ -n "${DELETE_OBJECTS}" ]; then
|
||||
aws s3api delete-objects \
|
||||
--bucket "${S3_BUCKET}" \
|
||||
--delete "${DELETE_OBJECTS}" \
|
||||
--region "${S3_REGION}" \
|
||||
>/dev/null 2>&1
|
||||
|
||||
echo "✓ Deleted ${DELETE_COUNT} old delta(s)"
|
||||
fi
|
||||
fi
|
||||
else
|
||||
echo "ℹ️ Only ${DELTA_COUNT} delta(s) exist, no cleanup needed"
|
||||
fi
|
||||
|
||||
# Cleanup delta tarball
|
||||
rm -f "${DELTA_TARBALL}"
|
||||
|
||||
# Cleanup: Unmount OverlayFS and remove workspace
|
||||
echo ""
|
||||
echo "Cleaning up..."
|
||||
|
||||
if mount | grep -q "${CACHE_WORKSPACE}/merged"; then
|
||||
sudo umount "${CACHE_WORKSPACE}/merged" || {
|
||||
echo "⚠️ Warning: Failed to unmount ${CACHE_WORKSPACE}/merged"
|
||||
echo "Attempting lazy unmount..."
|
||||
sudo umount -l "${CACHE_WORKSPACE}/merged" || true
|
||||
}
|
||||
fi
|
||||
|
||||
# Remove workspace
|
||||
rm -rf "${CACHE_WORKSPACE}"
|
||||
fi
|
||||
|
||||
# Remove from registry
|
||||
if [ -f "${MOUNT_REGISTRY}" ]; then
|
||||
grep -v "^${TARGET_PATH}:" "${MOUNT_REGISTRY}" > "${MOUNT_REGISTRY}.tmp" 2>/dev/null || true
|
||||
mv "${MOUNT_REGISTRY}.tmp" "${MOUNT_REGISTRY}" 2>/dev/null || true
|
||||
fi
|
||||
|
||||
echo "✓ Cleanup completed"
|
||||
# Cleanup
|
||||
rm -f "${BASE_TARBALL}"
|
||||
|
||||
echo ""
|
||||
echo "=========================================="
|
||||
echo "Cache save completed successfully"
|
||||
echo "Mode: ${CACHE_MODE}"
|
||||
echo "Cache key: ${PRIMARY_KEY}"
|
||||
if [ -n "${DELTA_SIZE:-}" ]; then
|
||||
echo "Delta size: ${DELTA_SIZE}"
|
||||
fi
|
||||
echo "Cache size: ${BASE_SIZE}"
|
||||
echo "Cache key: ${CACHE_KEY}"
|
||||
echo "=========================================="
|
||||
|
||||
Reference in New Issue
Block a user