diff options
Diffstat (limited to 'scripts/zfs-replicate')
| -rwxr-xr-x | scripts/zfs-replicate | 25 |
1 files changed, 21 insertions, 4 deletions
diff --git a/scripts/zfs-replicate b/scripts/zfs-replicate index 02ffcf5..a4cb15d 100755 --- a/scripts/zfs-replicate +++ b/scripts/zfs-replicate @@ -25,9 +25,14 @@ YELLOW='\033[1;33m' RED='\033[0;31m' NC='\033[0m' -info() { echo -e "${GREEN}[INFO]${NC} $1"; } -warn() { echo -e "${YELLOW}[WARN]${NC} $1"; } -error() { echo -e "${RED}[ERROR]${NC} $1"; exit 1; } +# Diagnostics go to stderr, never stdout. determine_host below runs inside a +# command substitution, so anything these print on stdout would be captured +# into TRUENAS_HOST instead of reaching the terminal or the journal -- which is +# exactly how an unreachable TrueNAS used to abort this script with exit 1 and +# no output at all. stderr also keeps the captured hostname clean. +info() { echo -e "${GREEN}[INFO]${NC} $1" >&2; } +warn() { echo -e "${YELLOW}[WARN]${NC} $1" >&2; } +error() { echo -e "${RED}[ERROR]${NC} $1" >&2; exit 1; } command -v syncoid >/dev/null 2>&1 || error "syncoid not found. Install sanoid package." @@ -58,6 +63,12 @@ fi info "Starting ZFS replication to $TRUENAS_HOST" echo "" +# A failed dataset must not stop the others -- but it must not be forgotten +# either. This runs as a systemd oneshot on a nightly timer, so the exit code +# is the only signal anyone sees; exiting 0 after every dataset failed made a +# backup that had not run in months look identical to a working one. +failed=0 + for dataset in $DATASETS; do dest="$TRUENAS_USER@$TRUENAS_HOST:$TRUENAS_POOL/$BACKUP_PATH/${dataset#zroot/}" info "Replicating $dataset -> $dest" @@ -66,8 +77,14 @@ for dataset in $DATASETS; do info " Success" else warn " Failed (will retry next run)" + failed=$((failed + 1)) fi - echo "" + echo "" >&2 done +if [ "$failed" -gt 0 ]; then + warn "Replication complete with $failed of $(echo "$DATASETS" | wc -w) datasets failed." + exit 1 +fi + info "Replication complete." |