4 # Copyright © 2019 František Kučera (Frantovo.cz, GlobalCode.info)
6 # This program is free software: you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation, either version 3 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful,
12 # but WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 # GNU General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program. If not, see <http://www.gnu.org/licenses/>.
20 # VCS Backup is a configuration for setting up a version control system mirrors.
21 # Currently Mercurial (Hg) and Git are supported.
23 # - mirrors remote repositories
24 # - creates Btrfs subvolume for each repository
25 # - does periodic pull to keep mirrors up to date
26 # - does periodic Btrfs snapshot to keep history (git push --force done on the remote repository will lead to modifications or deletions in our current mirror, but previous versions will be kept in the snapshots)
27 # - provides web interface for remote clonning of our mirrors (see systemd and etc folders)
28 # - can be controlled over SSH by a sane person / owner of the system
29 # - provides reports in the recfile format (to be processed using GNU Recutils or Relational pipes):
30 # - list of repositories/mirrors
31 # - results of pull operations
34 # This is an asynchronous message-driven shell script that runs distributed across two machines and four user accounts. You have been warned :-)
37 # Server-side configuration: / loadServerConfigFile()
38 VCS_BACKUP_DATA_DIR="/mnt/data";
39 VCS_BACKUP_CURRENT_DIR="$VCS_BACKUP_DATA_DIR/current";
40 VCS_BACKUP_PUBLIC_DIR="$VCS_BACKUP_DATA_DIR/public";
41 VCS_BACKUP_CONFIG_DIR="$VCS_BACKUP_DATA_DIR/config";
42 VCS_BACKUP_SNAPSHOT_DIR="$VCS_BACKUP_DATA_DIR/snapshot";
43 VCS_BACKUP_SUBVOLUME_SOCKET="/run/vcs-backup-subvolume";
44 VCS_BACKUP_CLONE_SOCKET="/run/vcs-backup-clone/socket"; # the directory will be writable by ${VCS_BACKUP_USER}
45 VCS_BACKUP_CLONE_CALLBACK_SOCKET="clone-callback";
46 VCS_BACKUP_USER="vcs-backup";
47 VCS_BACKUP_MANAGER="vcs-backup-manager";
49 # Client-side configuration: / see loadClientConfigFile()
50 VCS_BACKUP_SERVER="$VCS_BACKUP_MANAGER@localhost";
51 VCS_BACKUP_SSH_COMMAND=(ssh "$VCS_BACKUP_SERVER");
53 # Installation – check and do it by hand:
54 # There should be already mounted Btrfs at $VCS_BACKUP_DATA_DIR
55 installInstructions() {
56 cp vcs-backup.sh /usr/local/bin/
57 adduser --disabled-password "$VCS_BACKUP"
58 adduser --disabled-password "$VCS_BACKUP_MANAGER"
60 mkdir "$VCS_BACKUP_CURRENT_DIR";
61 mkdir "$VCS_BACKUP_CONFIG_DIR";
62 mkdir "$VCS_BACKUP_SNAPSHOT_DIR";
63 mkdir "$(dirname VCS_BACKUP_CLONE_SOCKET)"
65 chown "${VCS_BACKUP_USER}:${VCS_BACKUP_USER}" "$(dirname VCS_BACKUP_CLONE_SOCKET)"
66 chown "${VCS_BACKUP_MANAGER}:${VCS_BACKUP_MANAGER}" "$VCS_BACKUP_CONFIG_DIR"
70 # --- Private functions: ---------------------------------------------------------------------------
73 # $1 = VCS type: hg, git
75 isValidTypeAndURL() { ([[ "$1" == "hg" || "$1" == "git" ]]) && [[ $(echo "$2" | wc -l) == 1 ]] && [[ $(echo "$2" | grep -E '^(http|https|ssh)://([a-zA-Z0-9_-][a-zA-Z0-9_-.]*/?)+$' | wc -l) == 1 ]]; }
78 # $1 = path to the config file
79 loadConfigFile() { if [ -f "$1" ]; then . "$1"; fi }
80 loadClientConfigFile() { loadConfigFile ~/.config/vcs-backup/client.cfg; }
81 loadServerConfigFile() { loadConfigFile "/etc/vcs-backup/server.cfg"; }
85 urlToRelativeDirectoryPath() {
86 echo "$1" | sed -E 's@^[^:]+://@@g';
90 # $1 = optional value (if missing, reads STDIN)
91 escapeRecfileValue() { if [[ $# = 0 ]]; then awk '{ if (NR > 1) { printf "+ " } print $_ }'; else echo "${@}" | ${FUNCNAME[0]}; fi }
96 printRecfileKeyValue() { echo -n "$1: "; escapeRecfileValue "$2"; }
98 # --- Public interface functions: ------------------------------------------------------------------
100 # Environment: client
101 # $1 = VCS type: hg, git
103 # $3 = "public" or "private" (default), whether the repository should be available through the public web interface
104 # $4 = "clone" (optional), if present, will also clone the backup locally
105 vcs_backup_public_clientSubmitBackupRequest() {
106 if isValidTypeAndURL "$1" "$2"; then
107 ${VCS_BACKUP_SSH_COMMAND[@]} vcs-backup.sh serverSubmitBackupRequest "$1" "$2" "$3" "$4"
108 if [[ "$4" == "clone" ]]; then
109 if [[ "$1" == "hg" ]]; then hg clone "ssh://${VCS_BACKUP_SERVER}/$VCS_BACKUP_CURRENT_DIR/$1/$(urlToRelativeDirectoryPath $2)";
110 elif [[ "$1" == "git" ]]; then git clone "ssh://${VCS_BACKUP_SERVER}/$VCS_BACKUP_CURRENT_DIR/$1/$(urlToRelativeDirectoryPath $2)";
114 echo "Unsupported VCS type: '$1' or URL: '$2'" >&2;
118 # Environment: server
119 # User: $VCS_BACKUP_MANAGER
120 # has same parameters as clientSubmitBackupRequest (see above)
121 vcs_backup_public_serverSubmitBackupRequest() {
122 if isValidTypeAndURL "$1" "$2"; then
123 relativePath=$1/$(urlToRelativeDirectoryPath "$2");
124 absolutePath="$VCS_BACKUP_CONFIG_DIR/$relativePath";
125 # TODO: stop if directory already exists / only add public link?
126 mkdir -p "$absolutePath";
127 echo "$2" > "$absolutePath/url.txt"
128 echo "submited" > "$absolutePath/state.txt"
129 setfacl -m u:${VCS_BACKUP_USER}:r "$absolutePath/url.txt"
130 setfacl -m u:${VCS_BACKUP_USER}:rw "$absolutePath/state.txt"
132 if [[ "$3" == "public" ]]; then
133 cd "$VCS_BACKUP_PUBLIC_DIR";
134 mkdir -p "$(dirname $relativePath)";
135 ln -rs "../current/$relativePath" "$(dirname $relativePath)";
138 if [[ "$4" == "clone" ]]; then
139 callBackSocket=$absolutePath/${VCS_BACKUP_CLONE_CALLBACK_SOCKET}
140 socat -u "unix-recvfrom:$callBackSocket,mode=777" - | while read m; do # TODO: ,group=${VCS_BACKUP_USER} and no 777 ?
141 echo "Message from the service: $m";
146 echo "$relativePath" | socat -u - unix-send:${VCS_BACKUP_SUBVOLUME_SOCKET};
148 if [[ "$4" == "clone" ]]; then
149 echo "Waiting for a message from the service on $callBackSocket (PID $callBackPID)";
150 wait -n $callBackPID;
153 echo "Unsupported VCS type: '$1' or URL: '$2'" >&2;
157 # Environment: server
159 # Should be started as a systemd/init service.
160 # - reads messages from from the subvolume socket – message contains the relative directory path
161 # - creates a subvolume for given repository + necesary parent directories
162 # - sends a message to the clone service → start cloning into the created subvolume
163 vcs_backup_public_serverStartSubvolumeService() {
164 socat -u "unix-recv:${VCS_BACKUP_SUBVOLUME_SOCKET},group=${VCS_BACKUP_MANAGER},mode=770" - | while read d; do
165 mkdir -p $(dirname "$VCS_BACKUP_CURRENT_DIR/$d");
166 if [[ -e "$VCS_BACKUP_CURRENT_DIR/$d" ]]; then
167 callBackSocket="$VCS_BACKUP_CONFIG_DIR/$d/$VCS_BACKUP_CLONE_CALLBACK_SOCKET";
168 if [[ -e "$callBackSocket" ]]; then
169 echo "alreadyDone" | socat -u - unix-send:"$callBackSocket";
172 btrfs subvolume create "$VCS_BACKUP_CURRENT_DIR/$d" && \
173 echo "subvolumeCreated" > "$VCS_BACKUP_CONFIG_DIR/$d/state.txt" && \
174 chown "${VCS_BACKUP_USER}:${VCS_BACKUP_USER}" "$VCS_BACKUP_CURRENT_DIR/$d" && \
175 echo "$d" | socat -u - unix-send:${VCS_BACKUP_CLONE_SOCKET};
180 # Environment: server
181 # User: $VCS_BACKUP_USER
182 # should be started as a systemd/init service
183 vcs_backup_public_serverStartCloneService() {
184 socat -u "unix-recv:${VCS_BACKUP_CLONE_SOCKET},mode=700" - | while read d; do
185 vcsType=$(echo "$d" | sed 's@/.*@@g');
186 url=$(cat "$VCS_BACKUP_CONFIG_DIR/$d/url.txt");
188 if isValidTypeAndURL "$vcsType" "$url"; then
189 if [[ "$vcsType" == "hg" ]]; then hg clone -U "$url" "$VCS_BACKUP_CURRENT_DIR/$d";
190 elif [[ "$vcsType" == "git" ]]; then git clone --mirror "$url" "$VCS_BACKUP_CURRENT_DIR/$d";
191 fi && echo "cloned" > "$VCS_BACKUP_CONFIG_DIR/$d/state.txt";
193 echo "Unsupported VCS type: '$vcsType' or URL: '$url'" >&2;
196 callBackSocket="$VCS_BACKUP_CONFIG_DIR/$d/$VCS_BACKUP_CLONE_CALLBACK_SOCKET";
197 if [[ -e "$callBackSocket" ]]; then
198 echo "done" | socat -u - unix-send:"$callBackSocket";
203 # Environment: client
204 # prints list of repositories in recfile format
205 # usage example: vcs-backup.sh clientListRepositories | relpipe-in-recfile | relpipe-out-tabular
206 vcs_backup_public_clientListRepositories() {
207 ${VCS_BACKUP_SSH_COMMAND[@]} vcs-backup.sh serverListRepositories;
210 # Environment: server
211 # User: $VCS_BACKUP_MANAGER
212 vcs_backup_public_serverListRepositories() {
213 printRecfileKeyValue "%rec" "repository";
214 printRecfileKeyValue "%type" "bytes int";
215 printRecfileKeyValue "%type" "public bool";
218 find "$VCS_BACKUP_CONFIG_DIR" -name url.txt -printf '%P\n' | sort | xargs dirname | while read d; do
219 url=$(cat "$VCS_BACKUP_CONFIG_DIR/$d/url.txt");
220 state=$(cat "$VCS_BACKUP_CONFIG_DIR/$d/state.txt");
221 vcsType=$(echo "$d" | sed 's@/.*@@g');
222 sizeBytes=$(du -sb "$VCS_BACKUP_CURRENT_DIR/$d" | cut -f1);
223 [[ -e "$VCS_BACKUP_PUBLIC_DIR/$d" ]] && public="true" || public="false";
225 if [[ "$vcsType" == "hg" ]]; then lastCommit=$(hg log --limit 1 --template '{date|isodatesec}' -R "$VCS_BACKUP_CURRENT_DIR/$d" 2>/dev/null);
226 elif [[ "$vcsType" == "git" ]]; then lastCommit=$(git -C "$VCS_BACKUP_CURRENT_DIR/$d" log --max-count=1 --pretty="%ai");
227 else lastCommit=""; fi
229 printRecfileKeyValue "type" "$vcsType";
230 printRecfileKeyValue "url" "$url";
231 printRecfileKeyValue "state" "$state";
232 printRecfileKeyValue "public" "$public";
233 printRecfileKeyValue "serverPath" "$VCS_BACKUP_CURRENT_DIR/$d";
234 printRecfileKeyValue "size" "$sizeBytes";
235 printRecfileKeyValue "lastCommit" "$lastCommit";
240 # Environment: server
241 # User: $VCS_BACKUP_USER
242 # should be called from cron (usually every day)
243 vcs_backup_public_serverPullCronTask() {
244 printRecfileKeyValue "%rec" "pull";
245 printRecfileKeyValue "%type" "started date";
246 printRecfileKeyValue "%type" "finished date";
247 printRecfileKeyValue "%type" "duration int";
248 printRecfileKeyValue "%type" "resultCode int";
250 find "$VCS_BACKUP_CONFIG_DIR" -name url.txt -printf '%P\n' | sort | xargs dirname | while read d; do
251 state=$(cat "$VCS_BACKUP_CONFIG_DIR/$d/state.txt");
252 vcsType=$(echo "$d" | sed 's@/.*@@g');
253 absolutePath="$VCS_BACKUP_CURRENT_DIR/$d";
255 pullStarted=$(date --iso-8601=s);
256 pullStartedMiliseconds=$(($(date +%s%N)/1000000));
261 if [[ "$state" == "cloned" ]]; then
262 if [[ "$vcsType" == "hg" ]]; then pullResult=$(hg pull --force --repository "$absolutePath" 2>&1); pullResultCode=$?;
263 elif [[ "$vcsType" == "git" ]]; then pullResult=$(git -C "$absolutePath" fetch 2>&1); pullResultCode=$?;
265 pullFinished=$(date --iso-8601=s);
266 pullFinishedMiliseconds=$(($(date +%s%N)/1000000));
267 pullDuration=$(( $pullFinishedMiliseconds - $pullStartedMiliseconds ));
270 printRecfileKeyValue "serverPath" "$absolutePath";
271 printRecfileKeyValue "type" "$vcsType";
272 printRecfileKeyValue "state" "$state";
273 printRecfileKeyValue "started" "$pullStarted";
274 printRecfileKeyValue "finished" "$pullFinished";
275 printRecfileKeyValue "duration" "$pullDuration";
276 printRecfileKeyValue "resultCode" "$pullResultCode";
277 printRecfileKeyValue "message" "$pullResult";
282 # Environment: server
284 # should be called from cron (usually every day) after Pull (see above)
285 vcs_backup_public_serverSnapshotCronTask() {
286 printRecfileKeyValue "%rec" "snapshot";
287 printRecfileKeyValue "%type" "started date";
288 printRecfileKeyValue "%type" "finished date";
289 printRecfileKeyValue "%type" "duration int";
290 printRecfileKeyValue "%type" "resultCode int";
292 find "$VCS_BACKUP_CONFIG_DIR" -name url.txt -printf '%P\n' | sort | xargs dirname | while read d; do
293 state=$(cat "$VCS_BACKUP_CONFIG_DIR/$d/state.txt");
294 vcsType=$(echo "$d" | sed 's@/.*@@g');
295 absolutePath="$VCS_BACKUP_CURRENT_DIR/$d";
297 started=$(date --iso-8601=s);
298 startedMiliseconds=$(($(date +%s%N)/1000000));
304 if [[ "$state" == "cloned" ]]; then
305 snapshotPath="$VCS_BACKUP_SNAPSHOT_DIR/$d/$(date --iso-8601=date)";
306 mkdir -p $(dirname "$snapshotPath");
307 result=$(btrfs subvolume snapshot -r "$absolutePath" "$snapshotPath" 2>&1);
309 finished=$(date --iso-8601=s);
310 finishedMiliseconds=$(($(date +%s%N)/1000000));
311 duration=$(( $finishedMiliseconds - $startedMiliseconds ));
314 printRecfileKeyValue "currentPath" "$absolutePath";
315 printRecfileKeyValue "snapshotPath" "$snapshotPath";
316 printRecfileKeyValue "type" "$vcsType";
317 printRecfileKeyValue "state" "$state";
318 printRecfileKeyValue "started" "$started";
319 printRecfileKeyValue "finished" "$finished";
320 printRecfileKeyValue "duration" "$duration";
321 printRecfileKeyValue "resultCode" "$resultCode";
322 printRecfileKeyValue "message" "$result";
327 # --- Single entry-point: --------------------------------------------------------------------------
329 loadClientConfigFile;
330 loadServerConfigFile;
331 PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin";
332 PUBLIC_FUNCTION_PREFIX="vcs_backup_public_";
333 if type -t "$PUBLIC_FUNCTION_PREFIX$1" > /dev/null; then
334 "$PUBLIC_FUNCTION_PREFIX${@:1}";
335 elif [[ $(basename $0) == "vcs-backup-clone-private-hg" ]]; then "${PUBLIC_FUNCTION_PREFIX}clientSubmitBackupRequest" hg "$1" private clone;
336 elif [[ $(basename $0) == "vcs-backup-clone-private-git" ]]; then "${PUBLIC_FUNCTION_PREFIX}clientSubmitBackupRequest" git "$1" private clone;
337 elif [[ $(basename $0) == "vcs-backup-clone-public-hg" ]]; then "${PUBLIC_FUNCTION_PREFIX}clientSubmitBackupRequest" hg "$1" public clone;
338 elif [[ $(basename $0) == "vcs-backup-clone-public-git" ]]; then "${PUBLIC_FUNCTION_PREFIX}clientSubmitBackupRequest" git "$1" public clone;
340 echo "Unsupported sub-command: $1" >&2
341 echo "Available sub-commands:" >&2
342 declare -F | grep "$PUBLIC_FUNCTION_PREFIX" | sed "s/.*$PUBLIC_FUNCTION_PREFIX/ /g" >&2