Bootstrap new installation with boot and main networks

Adaptation for the bootstrap command to run a new installation with two
network.
- Boot network. It's required to perform provisioning process of new
  node. Also it's a private isolate network allows to transfer a
  private keys for agents.
- Main network. It's a base network for communication.

Signed-off-by: Anton Kremenetsky <anton.kremenetsky@gmail.com>

Author: akremenetsky

genesis_devtools/cmd/cli.py

@@ -45,6 +45,7 @@
 BOOTSTRAP_TAG = "bootstrap"
 LaunchModeType = tp.Literal["core", "element", "custom"]
 GC_CIDR = ipaddress.IPv4Network("10.20.0.0/22")
+GC_BOOT_CIDR = ipaddress.IPv4Network("10.30.0.0/24")
 
 
 @click.group(invoke_without_command=True)
@@ -498,14 +499,32 @@ def push_cmd(
 @click.option(
     "--cidr",
     default=GC_CIDR,
-    help="Network CIDR",
+    help="The main network CIDR",
     show_default=True,
     type=ipaddress.IPv4Network,
 )
 @click.option(
     "--bridge",
     default=None,
-    help="Name of the linux bridge, it will be created if not set.",
+    help=(
+        "Name of the linux bridge for the main network, "
+        "it will be created if not set."
+    ),
+)
+@click.option(
+    "--boot-cidr",
+    default=GC_BOOT_CIDR,
+    help="The bootstrap network CIDR",
+    show_default=True,
+    type=ipaddress.IPv4Network,
+)
+@click.option(
+    "--boot-bridge",
+    default=None,
+    help=(
+        "Name of the linux bridge for the bootstrap network, "
+        "it will be created if not set."
+    ),
 )
 @click.option(
     "-f",
@@ -607,6 +626,8 @@ def bootstrap_cmd(
     stand_spec: str | None,
     cidr: ipaddress.IPv4Network,
     bridge: str | None,
+    boot_cidr: ipaddress.IPv4Network,
+    boot_bridge: str | None,
     force: bool,
     no_wait: bool,
     use_image_inplace: bool,
@@ -676,6 +697,8 @@ def bootstrap_cmd(
             stand_spec=stand_spec,
             cidr=cidr,
             bridge=bridge,
+            boot_cidr=boot_cidr,
+            boot_bridge=boot_bridge,
             force=force,
             use_image_inplace=use_image_inplace,
             repository=repository,
@@ -1305,6 +1328,8 @@ def _bootstrap_core(
     stand_spec: tp.Dict[str, tp.Any] | None,
     cidr: ipaddress.IPv4Network,
     bridge: str | None,
+    boot_cidr: ipaddress.IPv4Network,
+    boot_bridge: str | None,
     force: bool,
     use_image_inplace: bool,
     repository: str,
@@ -1314,11 +1339,17 @@ def _bootstrap_core(
     logger.info("Starting genesis bootstrap in 'core' mode")
 
     net_name = utils.installation_net_name(name)
-    default_stand_network = stand_models.Network(
+    default_stand_main_network = stand_models.Network(
         name=bridge if bridge else net_name,
         cidr=cidr,
         managed_network=False if bridge else True,
     )
+    boot_net_name = utils.installation_boot_net_name(name)
+    default_stand_boot_network = stand_models.Network(
+        name=boot_bridge if boot_bridge else boot_net_name,
+        cidr=boot_cidr,
+        managed_network=False if boot_bridge else True,
+    )
 
     # Single bootstrap stand
     if stand_spec is None:
@@ -1330,14 +1361,18 @@ def _bootstrap_core(
             use_image_inplace=use_image_inplace,
             cores=cores,
             memory=memory,
-            network=default_stand_network,
+            network=default_stand_main_network,
+            boot_network=default_stand_boot_network,
             bootstrap_name=bootstrap_domain_name,
             hypervisors=hypervisors,
         )
     else:
         dev_stand = stand_models.Stand.from_spec(stand_spec)
         if dev_stand.network.is_dummy:
-            dev_stand.network = default_stand_network
+            dev_stand.network = default_stand_main_network
+
+        if dev_stand.boot_network.is_dummy:
+            dev_stand.boot_network = default_stand_boot_network
 
         # Assign the image to bootstraps if it wasn't specified
         # in the specification.
@@ -1368,8 +1403,13 @@ def _bootstrap_core(
         infra.delete_stand(stand)
         logger.info(f"Destroyed old genesis installation: {dev_stand.name}")
 
-    infra.create_stand(dev_stand, repository=repository)
-    logger.info("Launched genesis installation")
+    try:
+        infra.create_stand(dev_stand, repository=repository)
+        logger.info("Launched genesis installation")
+    except Exception:
+        infra.delete_stand(dev_stand)
+        logger.error(f"Failed to launch genesis installation {dev_stand.name}")
+        raise
 
     cidr = dev_stand.network.cidr
     logger.important(

genesis_devtools/infra/driver/libvirt.py

@@ -70,6 +70,24 @@ def _extract_net_from_bootstrap(
             name=name, cidr=cidr, managed_network=managed_network, dhcp=dhcp
         )
 
+    def _extract_boot_net_from_bootstrap(
+        self, bootstrap: minidom.Document
+    ) -> models.Network:
+        try:
+            net = bootstrap.getElementsByTagName(vc.GENESIS_META_BOOT_NET_TAG)[
+                0
+            ]
+        except Exception:
+            return models.Network.dummy()
+
+        name = net.firstChild.nodeValue
+        cidr = ipaddress.IPv4Network(net.getAttribute("cidr"))
+        managed_network = bool(int(net.getAttribute("managed_network")))
+        dhcp = bool(int(net.getAttribute("dhcp")))
+        return models.Network(
+            name=name, cidr=cidr, managed_network=managed_network, dhcp=dhcp
+        )
+
     def _domain2bootstrap(self, domain: minidom.Document) -> models.Bootstrap:
         node = self._domain2node(domain)
         return models.Bootstrap.from_node(node)
@@ -111,6 +129,9 @@ def list_stands(self) -> tp.List[models.Stand]:
             if node_type == "bootstrap":
                 stand.bootstraps.append(self._domain2bootstrap(domain))
                 stand.network = self._extract_net_from_bootstrap(domain)
+                stand.boot_network = self._extract_boot_net_from_bootstrap(
+                    domain
+                )
             elif node_type == "baremetal":
                 stand.baremetals.append(self._domain2node(domain))
             else:
@@ -136,19 +157,35 @@ def create_stand(
         ):
             raise ValueError(f"Some domain in stand {stand} already exists")
 
-        if stand.network.managed_network and libvirt.has_net(stand.network):
-            raise ValueError(f"Network {stand.network} already exists")
+        if stand.network.managed_network and libvirt.has_net(
+            stand.network.name
+        ):
+            raise ValueError(f"Network {stand.network.name} already exists")
+
+        if stand.boot_network.managed_network and libvirt.has_net(
+            stand.boot_network.name
+        ):
+            raise ValueError(
+                f"Network {stand.boot_network.name} already exists"
+            )
 
+        # Main network for ordinary communication
         if stand.network.managed_network:
             libvirt.create_nat_network(
                 name=stand.network.name,
                 cidr=stand.network.cidr,
                 dhcp_enabled=stand.network.dhcp,
             )
 
+        # Isolated network for bootstrap procedure
+        if stand.boot_network.managed_network:
+            libvirt.create_isolated_network(
+                name=stand.boot_network.name,
+            )
+
         # Prepare config drive for the bootstrap node
         spec = {
-            "schema_version": 1,
+            "schema_version": 2,
             "stand": dataclasses.asdict(stand),
             **extra_data,
         }
@@ -181,6 +218,17 @@ def create_stand(
                         "dhcp": int(stand.network.dhcp),
                     },
                 ),
+                self._tag(
+                    vc.GENESIS_META_BOOT_NET_TAG,
+                    stand.boot_network.name,
+                    {
+                        "cidr": str(stand.boot_network.cidr),
+                        "managed_network": int(
+                            stand.boot_network.managed_network
+                        ),
+                        "dhcp": int(stand.boot_network.dhcp),
+                    },
+                ),
             )
 
             libvirt.create_domain(
@@ -190,10 +238,7 @@ def create_stand(
                 cores=bootstrap.cores,
                 memory=bootstrap.memory,
                 disks=bootstrap.disks,
-                network=stand.network.name,
-                net_type=(
-                    "network" if stand.network.managed_network else "bridge"
-                ),
+                networks=(stand.network, stand.boot_network),
                 meta_tags=tags,
                 config_drive=config_drive_path,
             )
@@ -214,10 +259,8 @@ def create_stand(
                 use_image_inplace=node.use_image_inplace,
                 cores=node.cores,
                 memory=node.memory,
-                network=stand.network.name,
-                net_type=(
-                    "network" if stand.network.managed_network else "bridge"
-                ),
+                # Put new node to the boot network
+                networks=(stand.boot_network,),
                 meta_tags=tags,
                 disks=node.disks,
                 boot="network",
@@ -226,7 +269,9 @@ def create_stand(
     def delete_stand(self, stand: models.Stand) -> None:
         """Delete the stand."""
         for node in itertools.chain(stand.bootstraps, stand.baremetals):
-            libvirt.destroy_domain(node.name)
+            if libvirt.has_domain(node.name):
+                libvirt.destroy_domain(node.name)
 
-        if stand.network.managed_network:
-            libvirt.destroy_net(stand.network.name)
+        for net in (stand.network.name, stand.boot_network.name):
+            if libvirt.has_net(net):
+                libvirt.destroy_net(net)

genesis_devtools/infra/libvirt/constants.py

@@ -23,5 +23,6 @@
 GENESIS_META_MEM_TAG = "genesis:mem"
 GENESIS_META_IMAGE_TAG = "genesis:image"
 GENESIS_META_NET_TAG = "genesis:network"
+GENESIS_META_BOOT_NET_TAG = "genesis:boot_network"
 
 BootMode = tp.Literal["hd", "network"]