salt: Schedule daily backup on bootstrap node

CHANGELOG.md

@@ -1,5 +1,12 @@
 # CHANGELOG
 ## Release 2.10.4 (in development)
+
+### Features Added
+
+- A daily backup of the bootstrap node is now automatically scheduled.
+  All the backups are also replicated onto all the master nodes.
+  (PR [#3557](https://github.com/scality/metalk8s/pull/3557))
+
 ## Release 2.10.3
 ### Enhancements
 

buildchain/buildchain/salt_tree.py

@@ -522,6 +522,8 @@ def _get_parts(self) -> Iterator[str]:
     Path("salt/metalk8s/orchestrate/deploy_node.sls"),
     Path("salt/metalk8s/orchestrate/etcd.sls"),
     Path("salt/metalk8s/orchestrate/register_etcd.sls"),
+    Path("salt/metalk8s/orchestrate/backup/files/job.yaml.j2"),
+    Path("salt/metalk8s/orchestrate/backup/replication.sls"),
     Path("salt/metalk8s/orchestrate/bootstrap/init.sls"),
     Path("salt/metalk8s/orchestrate/bootstrap/accept-minion.sls"),
     Path("salt/metalk8s/orchestrate/bootstrap/pre-downgrade.sls"),

docs/operation/disaster_recovery/bootstrap_backup_restore.rst

@@ -25,7 +25,7 @@ To create a new backup file, run the following command:
 
     /srv/scality/metalk8s-|version|/backup.sh
 
-Backup archives are stored in /var/lib/metalk8s/backups.
+Backup archives are stored in /var/lib/metalk8s/backups on all master nodes.
 
 Restoring a Bootstrap Node
 **************************

salt/_modules/metalk8s.py

@@ -691,3 +691,27 @@ def configure_archive(archive, remove=False):
     msg = "Archive '{0}' {1}".format(archive, msg)
     log.info(msg)
     return msg
+
+
+def backup_node():
+    metalk8s_version = __pillar__["metalk8s"]["cluster_version"]
+    archives = get_archives()
+
+    try:
+        archive_path = archives[f"metalk8s-{metalk8s_version}"]["path"]
+    except KeyError as exc:
+        raise CommandExecutionError(
+            f"No MetalK8s archive found for version {metalk8s_version}"
+        ) from exc
+
+    backup_script = f"{archive_path}/backup.sh"
+    result = __salt__["cmd.run_all"](cmd=backup_script)
+    log.debug("Result: %r", result)
+
+    if result["retcode"] != 0:
+        output = result.get("stderr") or result["stdout"]
+        raise CommandExecutionError(f"Failed to run {backup_script}: {output}")
+
+    msg = "Backup successfully generated"
+    log.info(msg)
+    return msg

salt/_utils/kubernetes_utils.py

@@ -296,13 +296,27 @@ def __init__(self, model, api_cls, name, method_names=None):
             name="api_service",
         ),
         # }}}
+        # /apis/batch/v1/ {{{
+        ("batch/v1", "Job"): KindInfo(
+            model=k8s_client.V1Job,
+            api_cls=k8s_client.BatchV1Api,
+            name="namespaced_job",
+        ),
+        # }}}
         # /apis/batch/v1beta1/ {{{
         ("batch/v1beta1", "CronJob"): KindInfo(
             model=k8s_client.V1beta1CronJob,
             api_cls=k8s_client.BatchV1beta1Api,
             name="namespaced_cron_job",
         ),
         # }}}
+        # /apis/networking.k8s.io/v1/ {{{
+        ("networking.k8s.io/v1", "NetworkPolicy"): KindInfo(
+            model=k8s_client.V1NetworkPolicy,
+            api_cls=k8s_client.NetworkingV1Api,
+            name="namespaced_network_policy",
+        ),
+        # }}}
         # /apis/networking.k8s.io/v1beta1/ {{{
         ("networking.k8s.io/v1beta1", "Ingress"): KindInfo(
             model=k8s_client.NetworkingV1beta1Ingress,

salt/metalk8s/backup/configured.sls

@@ -0,0 +1,4 @@
+Schedule daily backup:
+  schedule.present:
+    - function: metalk8s.backup_node
+    - seconds: 86400

salt/metalk8s/orchestrate/backup/files/job.yaml.j2

@@ -0,0 +1,59 @@
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: backup-replication-{{ node }}
+  namespace: kube-system
+  labels:
+    app.kubernetes.io/name: backup-replication
+    app.kubernetes.io/part-of: metalk8s
+    app.kubernetes.io/managed-by: salt
+spec:
+  backoffLimit: 4
+  parallelism: 1
+  ttlSecondsAfterFinished: 120
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: backup-replication
+        app.kubernetes.io/part-of: metalk8s
+        app.kubernetes.io/managed-by: salt
+    spec:
+      nodeName: {{ node }}
+      containers:
+      - name: backup-replication
+        image: {{ image }}
+        command:
+          - wget
+          - --accept="*.tar.gz"
+          - --no-host-directories
+          - --mirror
+          - --ca-certificate=/certificate/ca.crt
+          - --user=$(BACKUP_USERNAME)
+          - --password=$(BACKUP_PASSWORD)
+          - https://backup/
+        env:
+          - name: BACKUP_USERNAME
+            valueFrom:
+              secretKeyRef:
+                name: backup-credentials
+                key: username
+          - name: BACKUP_PASSWORD
+            valueFrom:
+              secretKeyRef:
+                name: backup-credentials
+                key: password
+        volumeMounts:
+        - name: backups
+          mountPath: /backups
+        - name: ca-cert
+          mountPath: /certificate
+        workingDir: /backups
+      restartPolicy: Never
+      volumes:
+        - name: backups
+          hostPath:
+            path: /var/lib/metalk8s/backups
+            type: DirectoryOrCreate
+        - name: ca-cert
+          configMap:
+            name: backup-ca-cert

salt/metalk8s/orchestrate/backup/replication.sls

@@ -0,0 +1,16 @@
+{%- from "metalk8s/repo/macro.sls" import build_image_name with context %}
+
+{%- set master_nodes = salt.metalk8s.minions_by_role('master') %}
+{%- set image = build_image_name("metalk8s-utils") %}
+
+{%- for node in master_nodes | sort %}
+
+Schedule backup replication Job on {{ node }}:
+  metalk8s_kubernetes.object_present:
+    - name: salt://{{ slspath }}/files/job.yaml.j2
+    - template: jinja
+    - defaults:
+        node: {{ node }}
+        image: {{ image }}
+
+{%- endfor %}

salt/metalk8s/roles/bootstrap/init.sls

@@ -3,3 +3,4 @@ include:
   - metalk8s.kubernetes.kubelet
   - metalk8s.salt.master
   - metalk8s.utils
+  - metalk8s.backup.configured

salt/tests/unit/formulas/config.yaml

@@ -292,6 +292,15 @@ metalk8s:
           mode: master
           architecture: compact
 
+    backup:
+      files:
+        job.yaml.j2:
+          _cases:
+            "Create job manifest for a node":
+              extra_context:
+                node: master-1
+                image: registry/some-image-name:tag
+
     register_etcd.sls:
       _cases:
         "Target a new master node": &orch_target_master_node

salt/tests/unit/modules/files/test_metalk8s.yaml

@@ -425,3 +425,18 @@ configure_archive:
     invalid_path: True
     raises: True
     result: Invalid archive path
+
+backup_node:
+  # 0. Ok - Backup succeed
+  - archives: &backup_node_archives
+      metalk8s-2.10.0:
+        path: /tmp
+    result: Backup successfully generated
+  # 1. Nok - Backup script failed
+  - archives: *backup_node_archives
+    raises: True
+    result: "Failed to run /tmp/backup.sh: Boom!"
+  # 2. Nok - No such archive for this MetalK8s version
+  - archives: {}
+    raises: True
+    result: No MetalK8s archive found for version 2.10.0

salt/tests/unit/modules/test_metalk8s.py

@@ -621,3 +621,26 @@ def test__read_bootstrap_config(self, raises=False, result=None):
                     metalk8s._read_bootstrap_config(),
                     "config",
                 )
+
+    @utils.parameterized_from_cases(YAML_TESTS_CASES["backup_node"])
+    def test_backup_node(self, result, version="2.10.0", archives=None, raises=False):
+        def _cmd_run_all(cmd):
+            ret = {"retcode": 0, "stdout": "OK", "stderr": "Boom!"}
+            if raises:
+                ret["retcode"] = 1
+            return ret
+
+        salt_dict = {"cmd.run_all": MagicMock(side_effect=_cmd_run_all)}
+        pillar_dict = {"metalk8s": {"cluster_version": version}}
+
+        with patch.dict(metalk8s.__salt__, salt_dict), patch.dict(
+            metalk8s.__pillar__, pillar_dict
+        ), patch("metalk8s.get_archives", MagicMock(return_value=archives or {})):
+            if raises:
+                self.assertRaisesRegex(
+                    CommandExecutionError,
+                    result,
+                    metalk8s.backup_node,
+                )
+            else:
+                self.assertEqual(metalk8s.backup_node(), result)

scripts/backup.sh.in

@@ -12,7 +12,8 @@ TAR_OPTS=(
     "--atime-preserve"
     "--preserve-permissions"
 )
-BACKUP_ARCHIVE="/var/lib/metalk8s/backup_$(date -u +%Y%m%d_%H%M%S).tar.gz"
+BACKUP_ARCHIVE="/var/lib/metalk8s/backups/$(date -u +%Y%m%d_%H%M%S).tar.gz"
+REPLICATION=1
 
 _usage() {
     echo "$(basename "$0") [options]"
@@ -26,22 +27,26 @@ while (( "$#" )); do
   case "$1" in
     -v|--verbose)
       VERBOSE=1
-      shift
       ;;
     -l|--log-file)
       LOGFILE="$2"
-      shift 2
+      shift
       ;;
     -b|--backup-file)
       BACKUP_ARCHIVE="$2"
-      shift 2
+      shift
+      ;;
+    # Disable the backup replication on other master nodes
+    -n|--no-replication)
+      REPLICATION=0
       ;;
     *) # unsupported flags
       echo "Error: Unsupported flag $1" >&2
       _usage
       exit 1
       ;;
   esac
+  shift
 done
 
 TMPFILES=$(mktemp -d)
@@ -62,69 +67,9 @@ cleanup() {
 
 trap cleanup EXIT
 
-run_quiet() {
-    local name=$1
-    shift 1
-
-    echo -n "> ${name}..."
-    local start
-    start=$(date +%s)
-    set +e
-    "$@" 2>&1 | tee -ia "${LOGFILE}" > "${TMPFILES}/out"
-    local RC=$?
-    set -e
-    local end
-    end=$(date +%s)
-
-    local duration=$(( end - start ))
-
-    if [ $RC -eq 0 ]; then
-        echo " done [${duration}s]"
-    else
-        echo " fail [${duration}s]"
-        cat >/dev/stderr << EOM
-
-Failure while running step '${name}'
-
-Command: $@
-
-Output:
-
-<< BEGIN >>
-EOM
-        cat "${TMPFILES}/out" > /dev/stderr
-
-        cat >/dev/stderr << EOM
-<< END >>
-
-This script will now exit
-
-EOM
-
-        exit 1
-    fi
-}
-
-run_verbose() {
-    local name=$1
-    shift 1
-
-    echo "> ${name}..."
-    "$@"
-}
-
-run() {
-    if [ "$VERBOSE" -eq 1 ]; then
-        run_verbose "${@}"
-    else
-        run_quiet "${@}"
-    fi
-}
-
-die() {
-    echo 1>&2 "$@"
-    return 1
-}
+BASE_DIR=$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")
+# shellcheck disable=SC1090
+. "$BASE_DIR"/common.sh
 
 _save_cp() {
     local -r src="$(readlink -f "$1")"
@@ -225,7 +170,16 @@ EOF
     tar "${TAR_OPTS[@]}" -C "$BACKUP_DIR" -cz -f "$BACKUP_ARCHIVE" ./
 }
 
+replicate_archives() {
+    salt_master_exec=(crictl exec -i "$(get_salt_container)")
+
+    "${salt_master_exec[@]}" salt-run --state-output=mixed state.orchestrate \
+	metalk8s.orchestrate.backup.replication \
+	saltenv=metalk8s-@@VERSION
+}
+
 run "Backing up MetalK8s configurations" backup_metalk8s_conf
 run "Backing up CAs certificates and keys" backup_cas
 run "Backing up etcd data" backup_etcd
 run "Creating backup archive '$BACKUP_ARCHIVE'" create_archive
+(( REPLICATION )) && run "Replicating backup archives on master nodes" replicate_archives

scripts/downgrade.sh.in

@@ -190,4 +190,7 @@ run "Downgrading bootstrap" downgrade_bootstrap
 run "Launching the downgrade" launch_downgrade
 run "Launching the post-downgrade" launch_post_downgrade
 
-"$BASE_DIR"/backup.sh
+# NOTE: We use --no-replication flag since the backup-server is not available
+# in the previous version of MetalK8s
+# This flag can safely be removed in MetalK8s 2.12.
+"$BASE_DIR"/backup.sh --no-replication

scripts/upgrade.sh.in

@@ -193,7 +193,10 @@ patch_kubesystem_namespace() {
 run "Performing Pre-Upgrade checks" precheck_upgrade
 [ $DRY_RUN -eq 1 ] && exit 0
 
-"$BASE_DIR"/backup.sh
+# NOTE: We use --no-replication flag since the backup-server is not available
+# in the previous version of MetalK8s.
+# This flag can safely be removed in MetalK8s 2.12.
+"$BASE_DIR"/backup.sh --no-replication
 
 run "Upgrading bootstrap" upgrade_bootstrap
 run "Setting cluster version to $DESTINATION_VERSION" patch_kubesystem_namespace