Merge branch 'master' into rn-0.15.0

clab/clab.go

@@ -151,45 +151,68 @@ func (c *CLab) GlobalRuntime() runtime.ContainerRuntime {
 	return c.Runtimes[c.globalRuntime]
 }
 
-func (c *CLab) CreateNodes(ctx context.Context, workers uint) {
+func (c *CLab) CreateNodes(ctx context.Context, maxWorkers uint, serialNodes map[string]struct{}) {
+
 	wg := new(sync.WaitGroup)
-	wg.Add(int(workers))
-	nodesChan := make(chan nodes.Node)
-	// start workers
-	for i := uint(0); i < workers; i++ {
-		go func(i uint) {
-			defer wg.Done()
-			for {
-				select {
-				case node, ok := <-nodesChan:
-					if node == nil || !ok {
-						log.Debugf("Worker %d terminating...", i)
-						return
-					}
-					log.Debugf("Worker %d received node: %+v", i, node.Config())
-					// PreDeploy
-					err := node.PreDeploy(c.Config.Name, c.Dir.LabCA, c.Dir.LabCARoot)
-					if err != nil {
-						log.Errorf("failed pre-deploy phase for node %q: %v", node.Config().ShortName, err)
-						continue
-					}
-					// Deploy
-					err = node.Deploy(ctx)
-					if err != nil {
-						log.Errorf("failed deploy phase for node %q: %v", node.Config().ShortName, err)
-						continue
-					}
-				case <-ctx.Done():
+
+	concurrentChan := make(chan nodes.Node)
+	serialChan := make(chan nodes.Node)
+
+	workerFunc := func(i uint, input chan nodes.Node, wg *sync.WaitGroup) {
+		defer wg.Done()
+		for {
+			select {
+			case node, ok := <-input:
+				if node == nil || !ok {
+					log.Debugf("Worker %d terminating...", i)
 					return
 				}
+				log.Debugf("Worker %d received node: %+v", i, node.Config())
+				// PreDeploy
+				err := node.PreDeploy(c.Config.Name, c.Dir.LabCA, c.Dir.LabCARoot)
+				if err != nil {
+					log.Errorf("failed pre-deploy phase for node %q: %v", node.Config().ShortName, err)
+					continue
+				}
+				// Deploy
+				err = node.Deploy(ctx)
+				if err != nil {
+					log.Errorf("failed deploy phase for node %q: %v", node.Config().ShortName, err)
+					continue
+				}
+			case <-ctx.Done():
+				return
 			}
-		}(i)
+		}
+	}
+
+	// start concurrent workers
+	wg.Add(int(maxWorkers))
+	// it's safe to not check if all nodes are serial because in that case
+	// maxWorkers will be 0
+	for i := uint(0); i < maxWorkers; i++ {
+		go workerFunc(i, concurrentChan, wg)
 	}
+
+	// start the serial worker
+	if len(serialNodes) > 0 {
+		wg.Add(1)
+		go workerFunc(maxWorkers, serialChan, wg)
+	}
+
+	// send nodes to workers
 	for _, n := range c.Nodes {
-		nodesChan <- n
+		if _, ok := serialNodes[n.Config().LongName]; ok {
+			serialChan <- n
+			continue
+		}
+		concurrentChan <- n
 	}
+
 	// close channel to terminate the workers
-	close(nodesChan)
+	close(concurrentChan)
+	close(serialChan)
+
 	// wait for all workers to finish
 	wg.Wait()
 }
@@ -248,35 +271,56 @@ func (c *CLab) CreateLinks(ctx context.Context, workers uint, postdeploy bool) {
 	wg.Wait()
 }
 
-func (c *CLab) DeleteNodes(ctx context.Context, workers uint, deleteCandidates map[string]nodes.Node) {
+func (c *CLab) DeleteNodes(ctx context.Context, workers uint, deleteCandidates map[string]nodes.Node, serialNodes map[string]struct{}) {
+
 	wg := new(sync.WaitGroup)
 
-	nodeChan := make(chan nodes.Node)
-	wg.Add(int(workers))
-	for i := uint(0); i < workers; i++ {
-		go func(i uint) {
-			defer wg.Done()
-			for {
-				select {
-				case n := <-nodeChan:
-					if n == nil {
-						log.Debugf("Worker %d terminating...", i)
-						return
-					}
-					err := n.Delete(ctx)
-					if err != nil {
-						log.Errorf("could not remove container %q: %v", n.Config().LongName, err)
-					}
-				case <-ctx.Done():
+	concurrentChan := make(chan nodes.Node)
+	serialChan := make(chan nodes.Node)
+
+	workerFunc := func(i uint, input chan nodes.Node, wg *sync.WaitGroup) {
+		defer wg.Done()
+		for {
+			select {
+			case n := <-input:
+				if n == nil {
+					log.Debugf("Worker %d terminating...", i)
 					return
 				}
+				err := n.Delete(ctx)
+				if err != nil {
+					log.Errorf("could not remove container %q: %v", n.Config().LongName, err)
+				}
+			case <-ctx.Done():
+				return
 			}
-		}(i)
+		}
 	}
-	for _, n := range deleteCandidates {
-		nodeChan <- n
+
+	// start concurrent workers
+	wg.Add(int(workers))
+	for i := uint(0); i < workers; i++ {
+		go workerFunc(i, concurrentChan, wg)
+	}
+
+	// start the serial worker
+	if len(serialNodes) > 0 {
+		wg.Add(1)
+		go workerFunc(workers, serialChan, wg)
 	}
-	close(nodeChan)
+
+	// send nodes to workers
+	for _, n := range c.Nodes {
+		if _, ok := serialNodes[n.Config().LongName]; ok {
+			serialChan <- n
+			continue
+		}
+		concurrentChan <- n
+	}
+
+	// close channel to terminate the workers
+	close(concurrentChan)
+	close(serialChan)
 
 	wg.Wait()
 

cmd/deploy.go

@@ -119,12 +119,17 @@ var deployCmd = &cobra.Command{
 			linkWorkers = maxWorkers
 		}
 
-		// Serializing ignite workers due to busy device error
-		if _, ok := c.Runtimes[runtime.IgniteRuntime]; ok {
-			nodeWorkers = 1
+		// a set of workers that do not support concurrency
+		serialNodes := make(map[string]struct{})
+		for _, n := range c.Nodes {
+			if n.GetRuntime().GetName() == runtime.IgniteRuntime {
+				serialNodes[n.Config().LongName] = struct{}{}
+				// decreasing the num of nodeworkers as they are used for concurrent nodes
+				nodeWorkers = nodeWorkers - 1
+			}
 		}
 
-		c.CreateNodes(ctx, nodeWorkers)
+		c.CreateNodes(ctx, nodeWorkers, serialNodes)
 		c.CreateLinks(ctx, linkWorkers, false)
 		log.Debug("containers created, retrieving state and IP addresses...")
 
@@ -142,16 +147,17 @@ var deployCmd = &cobra.Command{
 			return err
 		}
 
-		var wg sync.WaitGroup
+		wg := &sync.WaitGroup{}
 		wg.Add(len(c.Nodes))
+
 		for _, node := range c.Nodes {
-			go func(node nodes.Node) {
+			go func(node nodes.Node, wg *sync.WaitGroup) {
 				defer wg.Done()
 				err := node.PostDeploy(ctx, c.Nodes)
 				if err != nil {
 					log.Errorf("failed to run postdeploy task for node %s: %v", node.Config().ShortName, err)
 				}
-			}(node)
+			}(node, wg)
 		}
 		wg.Wait()
 

cmd/destroy.go

@@ -190,13 +190,23 @@ func destroyLab(ctx context.Context, c *clab.CLab) (err error) {
 		maxWorkers = uint(len(c.Nodes))
 	}
 
+	// a set of workers that do not support concurrency
+	serialNodes := make(map[string]struct{})
+	for _, n := range c.Nodes {
+		if n.GetRuntime().GetName() == runtime.IgniteRuntime {
+			serialNodes[n.Config().LongName] = struct{}{}
+			// decreasing the num of maxWorkers as they are used for concurrent nodes
+			maxWorkers = maxWorkers - 1
+		}
+	}
+
 	// Serializing ignite workers due to busy device error
 	if _, ok := c.Runtimes[runtime.IgniteRuntime]; ok {
 		maxWorkers = 1
 	}
 
 	log.Infof("Destroying lab: %s", c.Config.Name)
-	c.DeleteNodes(ctx, maxWorkers, c.Nodes)
+	c.DeleteNodes(ctx, maxWorkers, c.Nodes, serialNodes)
 
 	// remove the lab directories
 	if cleanup {

docs/cmd/deploy.md

@@ -31,6 +31,15 @@ Refer to the [configuration artifacts](../manual/conf-artifacts.md) page to get
 #### max-workers
 With `--max-workers` flag it is possible to limit the amout of concurrent workers that create containers or wire virtual links. By default the number of workers equals the number of nodes/links to create.
 
+#### runtime
+Containerlab nodes can be started by different runtimes, with `docker` being the default one. Besides `docker`, containerlab has experimental support for `containerd` and `ignite` runtimes.
+
+A global runtime can be selected with a global `--runtime | -r` flag that will select a runtime to use. The supported value are:
+
+* `docker` - default
+* `containerd`
+* `ignite`
+
 ### Examples
 
 ```bash

docs/lab-examples/cvx03.md

@@ -3,7 +3,7 @@
 | **Description**               | Cumulus Linux Test Drive                                                                 |
 | **Components**                | [Cumulus Linux][cvx]                                                                     |
 | **Resource requirements**[^1] | :fontawesome-solid-microchip: 2 <br/>:fontawesome-solid-memory: 2 GB                     |
-| **Topology file**             | [lab-start.yml][topofile] <br/>[lab-final.yml][finalfile]                                                                     |
+| **Topology file**             | [lab-start.clab.yml][topofile] <br/>[lab-final.clab.yml][finalfile]                                                                     |
 | **Name**                      | cvx03                                                                                    |
 | **Version information**[^2]   | `cvx:4.3.0` `docker-ce:19.03.13`                                                         |
 
@@ -19,13 +19,13 @@ This is a "Cumulus Test Drive" topology designed to provide an overview of NVIDI
 2. BGP Unnumbered (Lab3) -- learn how to configure BGP unnumbered between leaf and spine switches and advertise locally connected interfaces.
 
 !!!note
-    Everything from Lab1 is already pre-configured when the topology is created with [lab-start.yml][topofile].
+    Everything from Lab1 is already pre-configured when the topology is created with [lab-start.clab.yml][topofile].
 
-Additionally, the lab directory contains a [lab-final.yml][finalfile] which will load final configurations as they appear at the end of Lab3.
+Additionally, the lab directory contains a [lab-final.clab.yml][finalfile] which will load final configurations as they appear at the end of Lab3.
 
 [cvx]: https://www.nvidia.com/en-gb/networking/ethernet-switching/cumulus-vx/
-[topofile]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx03/lab-start.yml
-[finalfile]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx03/lab-final.yml
+[topofile]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx03/lab-start.clab.yml
+[finalfile]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx03/lab-final.clab.yml
 
 [^1]: Resource requirements are provisional. Consult with the installation guides for additional information.
 [^2]: The lab has been validated using these versions of the required tools/components. Using versions other than stated might lead to a non-operational setup process.

docs/lab-examples/cvx04.md

@@ -3,7 +3,7 @@
 | **Description**               | Cumulus In The Cloud                                                                 |
 | **Components**                | [Cumulus Linux][cvx]                                                                     |
 | **Resource requirements**[^1] | :fontawesome-solid-microchip: 2 <br/>:fontawesome-solid-memory: 4 GB                     |
-| **Topology file**             | [symm-mh.yml][topo-mh] <br/>[symm-mlag.yml][topo-mlag]                                                                     |
+| **Topology file**             | [symm-mh.clab.yml][topo-mh] <br/>[symm-mlag.clab.yml][topo-mlag]                                                                     |
 | **Name**                      | cvx04                                                                                    |
 | **Version information**[^2]   | `cvx:4.3.0` `docker-ce:19.03.13`                                                         |
 
@@ -21,7 +21,7 @@ This is a "Cumulus In The Cloud" topology designed to demonstrate some of the ad
 
 ## Instructions
 
-Each configuration option is provided in its own configuration file -- [`symm-mh.yml`][topo-mh] or [`symm-mlag.yml`][topo-mlag]. See [instructions](/lab-examples/lab-examples/#how-to-deploy-a-lab-from-the-lab-catalog) for how to deploy a topology. 
+Each configuration option is provided in its own configuration file -- [`symm-mh.clab.yml`][topo-mh] or [`symm-mlag.clab.yml`][topo-mlag]. See [instructions](/lab-examples/lab-examples/#how-to-deploy-a-lab-from-the-lab-catalog) for how to deploy a topology. 
 
 Once up, each node can be accessed via ssh using its hostname (automatically populated in your `/etc/hosts` file) and the default credentials `root/root`:
 
@@ -34,13 +34,10 @@ Last login: Fri Jul  9 13:35:48 2021 from 192.168.223.1
 root@94992c82719f1172:mgmt:~# 
 ```
 
-!!!note
-    Due to the different boot order inside a container, BGPd may come up stuck waiting for IPv6 LLA of the peer. This issue only appears on the initial boot and can be fixed with the `vtysh -c 'clear ip bgp *` command.
-
 
 [cvx]: https://www.nvidia.com/en-gb/networking/ethernet-switching/cumulus-vx/
-[topo-mh]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx04/symm-mh.yml
-[topo-mlag]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx04/symm-mlag.yml
+[topo-mh]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx04/symm-mh.clab.yml
+[topo-mlag]: https://github.com/srl-labs/containerlab/tree/master/lab-examples/cvx04/symm-mlag.clab.yml
 
 [^1]: Resource requirements are provisional. Consult with the installation guides for additional information.
 [^2]: The lab has been validated using these versions of the required tools/components. Using versions other than stated might lead to a non-operational setup process.

docs/manual/kinds/cvx.md

@@ -12,7 +12,7 @@ CVX nodes launched with containerlab comes up with:
 CVX supports two modes of operation:
 
 * Using Firecracker micro-VMs -- this mode runs Cumulus VX inside a micro-VM on top of the native Cumulus kernel. This is the default way of running CVX nodes.
-* Using only the container runtime -- this mode runs Cumulus VX container image directly inside the container runtime (e.g. Docker). Due to the lack of Cumulus VX kernel modules, some features are not supported, most notable one being MLAG. In order to use this mode add `runtime: docker` under the cvx node definition (see also [this example](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx02/topo.yml)).
+* Using only the container runtime -- this mode runs Cumulus VX container image directly inside the container runtime (e.g. Docker). Due to the lack of Cumulus VX kernel modules, some features are not supported, most notable one being MLAG. In order to use this mode add `runtime: docker` under the cvx node definition (see also [this example](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx02/topo.clab.yml)).
 
 
 ## Managing cvx nodes
@@ -63,8 +63,8 @@ This will be addressed in the future releases.
 ## Lab examples
 The following labs feature CVX node:
 
-- [Cumulus Linux Test Drive](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx03/lab-start.yml)
-- [Cumulus and FRR](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx01/topo.yml)
+- [Cumulus Linux Test Drive](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx03/lab-start.clab.yml)
+- [Cumulus and FRR](https://github.com/srl-labs/containerlab/blob/master/lab-examples/cvx01/topo.clab.yml)
 
 
 ## Known issues or limitations

docs/manual/nodes.md

@@ -285,4 +285,23 @@ my-node:
   network-mode: host
 ```
 
-The `network-mode` configuration option set to `host` will launch the node in the [host networking mode](https://docs.docker.com/network/host/).
+The `network-mode` configuration option set to `host` will launch the node in the [host networking mode](https://docs.docker.com/network/host/).
+
+### runtime
+By default containerlab nodes will be started by `docker` container runtime. Besides that, containerlab has experimental support for `containerd` and `ignite` runtimes.
+
+It is possible to specify a global runtime with a global `--runtime` flag, or set the runtime on a per-node basis:
+
+Options for the runtime parameter are:
+- `docker`
+- `containerd`
+- `ignite`
+
+The default runtime can also be influenced via the `CLAB_RUNTIME` environment variable, which takes the same values as mentioned above.
+
+```yaml
+# example node definition with per-node runtime definition
+my-node:
+  image: alpine:3
+  runtime: containerd
+```