feat: add synchronization statistics logging

- Initialize statistics logging in the mirror script
- Update logging to include total, processed, skipped, success, and failed repositories
- Enhance summary report with detailed statistics after synchronization

This commit introduces a comprehensive logging mechanism for the
repository synchronization process. It tracks various metrics such as
total repositories, processed repositories, and their success or failure
status. This enhancement provides better visibility into the
synchronization process and helps in debugging and monitoring.
This commit is contained in:
songtianlun
2025-01-03 17:26:32 +08:00
parent f7770955a9
commit ca6a80da97
3 changed files with 123 additions and 22 deletions

107
mirror.sh
View File

@@ -8,45 +8,101 @@ GITEA_USER="$4"
GITEA_TOKEN="$5"
WORK_DIR="$6"
SKIP_REPOS="$7"
STATS_FILE="$8"
# 初始化统计
init_stats() {
cat > "$STATS_FILE" << EOF
{
"total_repos": 0,
"processed": 0,
"skipped": 0,
"success": 0,
"failed": 0,
"start_time": "$(date '+%Y-%m-%d %H:%M:%S')",
"end_time": "",
"details": {
"skipped_repos": [],
"success_repos": [],
"failed_repos": []
}
}
EOF
}
# 更新统计信息
update_stats() {
local key="$1"
local value="$2"
local type="$3" # 可以是 number 或 array
if [ "$type" = "array" ]; then
# 添加到数组
jq --arg key "$key" --arg value "$value" \
".details[$key] += [\$value]" "$STATS_FILE" > "$STATS_FILE.tmp"
else
# 更新数值
jq --arg key "$key" --arg value "$value" \
".[$key] = ($value | tonumber)" "$STATS_FILE" > "$STATS_FILE.tmp"
fi
mv "$STATS_FILE.tmp" "$STATS_FILE"
}
# 同步单个仓库
sync_repository() {
local repo="$1"
local success=true
# 检查 Gitea 仓库是否存在
if ! curl -s -o /dev/null -f -H "Authorization: token $GITEA_TOKEN" \
"$GITEA_URL/api/v1/repos/$GITEA_USER/$repo"; then
echo "在 Gitea 上创建仓库 $repo"
curl -X POST -H "Authorization: token $GITEA_TOKEN" \
if ! curl -X POST -H "Authorization: token $GITEA_TOKEN" \
-H "Content-Type: application/json" \
-d "{\"name\":\"$repo\",\"private\":false}" \
"$GITEA_URL/api/v1/user/repos"
"$GITEA_URL/api/v1/user/repos"; then
success=false
fi
fi
# 克隆和推送
[ -d "$repo" ] && rm -rf "$repo"
git clone --mirror "https://${GITHUB_TOKEN:+$GITHUB_TOKEN@}github.com/$GITHUB_USER/$repo.git" "$repo"
cd "$repo"
if ! git clone --mirror "https://${GITHUB_TOKEN:+$GITHUB_TOKEN@}github.com/$GITHUB_USER/$repo.git" "$repo"; then
success=false
else
cd "$repo"
# 尝试 mirror 推送
if ! git push --mirror "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git"; then
echo "mirror 推送失败,尝试逐个分支推送..."
# 尝试 mirror 推送
if ! git push --mirror "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git"; then
echo "mirror 推送失败,尝试逐个分支推送..."
# 获取所有分支
git fetch --all
# 获取所有分支
git fetch --all
# 推送每个分支
git for-each-ref --format='%(refname:short)' refs/heads/ | while read branch; do
echo "推送分支: $branch"
git push "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git" "$branch:$branch"
done
# 推送每个分支
git for-each-ref --format='%(refname:short)' refs/heads/ | while read branch; do
echo "推送分支: $branch"
if ! git push "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git" "$branch:$branch"; then
success=false
fi
done
# 推送所有标签
git push "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git" --tags
# 推送所有标签
if ! git push "https://$GITEA_USER:$GITEA_TOKEN@${GITEA_URL#https://}/$GITEA_USER/$repo.git" --tags; then
success=false
fi
fi
cd ..
fi
cd ..
# 更新统计
if [ "$success" = true ]; then
update_stats "success" "$(( $(jq '.success' "$STATS_FILE") + 1 ))" "number"
update_stats "success_repos" "$repo" "array"
else
update_stats "failed" "$(( $(jq '.failed' "$STATS_FILE") + 1 ))" "number"
update_stats "failed_repos" "$repo" "array"
fi
}
# 主同步逻辑
@@ -54,22 +110,37 @@ main() {
mkdir -p "$WORK_DIR"
cd "$WORK_DIR"
# 初始化统计
init_stats
# 获取仓库列表
repos=$(curl -s -H "Authorization: token $GITHUB_TOKEN" \
"https://api.github.com/user/repos?per_page=100&type=all" | \
jq -r '.[].name')
# 更新总仓库数
total_repos=$(echo "$repos" | wc -l)
update_stats "total_repos" "$total_repos" "number"
# 同步每个仓库
for repo in $repos; do
# 检查是否跳过
if echo "$SKIP_REPOS" | grep -q "$repo"; then
echo "跳过仓库: $repo"
update_stats "skipped" "$(( $(jq '.skipped' "$STATS_FILE") + 1 ))" "number"
update_stats "skipped_repos" "$repo" "array"
continue
fi
echo "处理仓库: $repo"
update_stats "processed" "$(( $(jq '.processed' "$STATS_FILE") + 1 ))" "number"
sync_repository "$repo"
done
# 更新结束时间
jq --arg time "$(date '+%Y-%m-%d %H:%M:%S')" \
'.end_time = $time' "$STATS_FILE" > "$STATS_FILE.tmp"
mv "$STATS_FILE.tmp" "$STATS_FILE"
}
main "$@"