Add filters for container state to docker input (#3950)

plugins/inputs/docker/README.md

@@ -31,6 +31,11 @@ to gather stats from the [Engine API](https://docs.docker.com/engine/api/v1.20/)
   container_name_include = []
   container_name_exclude = []
 
+  ## Container states to include and exclude. Globs accepted.
+  ## When empty only containers in the "running" state will be captured.
+  # container_state_include = []
+  # container_state_exclude = []
+
   ## Timeout for docker list, info, and stats commands
   timeout = "5s"
 

plugins/inputs/docker/docker.go

@@ -15,6 +15,7 @@ import (
 	"time"
 
 	"github.com/docker/docker/api/types"
+	"github.com/docker/docker/api/types/filters"
 	"github.com/docker/docker/api/types/swarm"
 	"github.com/influxdata/telegraf"
 	"github.com/influxdata/telegraf/filter"
@@ -25,7 +26,7 @@ import (
 // Docker object
 type Docker struct {
 	Endpoint       string
-	ContainerNames []string
+	ContainerNames []string // deprecated in 1.4; use container_name_include
 
 	GatherServices bool `toml:"gather_services"`
 
@@ -39,6 +40,9 @@ type Docker struct {
 	ContainerInclude []string `toml:"container_name_include"`
 	ContainerExclude []string `toml:"container_name_exclude"`
 
+	ContainerStateInclude []string `toml:"container_state_include"`
+	ContainerStateExclude []string `toml:"container_state_exclude"`
+
 	SSLCA              string `toml:"ssl_ca"`
 	SSLCert            string `toml:"ssl_cert"`
 	SSLKey             string `toml:"ssl_key"`
@@ -53,6 +57,7 @@ type Docker struct {
 	filtersCreated  bool
 	labelFilter     filter.Filter
 	containerFilter filter.Filter
+	stateFilter     filter.Filter
 }
 
 // KB, MB, GB, TB, PB...human friendly
@@ -67,7 +72,8 @@ const (
 )
 
 var (
-	sizeRegex = regexp.MustCompile(`^(\d+(\.\d+)*) ?([kKmMgGtTpP])?[bB]?$`)
+	sizeRegex       = regexp.MustCompile(`^(\d+(\.\d+)*) ?([kKmMgGtTpP])?[bB]?$`)
+	containerStates = []string{"created", "restarting", "running", "removing", "paused", "exited", "dead"}
 )
 
 var sampleConfig = `
@@ -87,6 +93,11 @@ var sampleConfig = `
   container_name_include = []
   container_name_exclude = []
 
+  ## Container states to include and exclude. Globs accepted.
+  ## When empty only containers in the "running" state will be captured.
+  # container_state_include = []
+  # container_state_exclude = []
+
   ## Timeout for docker list, info, and stats commands
   timeout = "5s"
 
@@ -148,6 +159,10 @@ func (d *Docker) Gather(acc telegraf.Accumulator) error {
 		if err != nil {
 			return err
 		}
+		err = d.createContainerStateFilters()
+		if err != nil {
+			return err
+		}
 		d.filtersCreated = true
 	}
 
@@ -164,8 +179,22 @@ func (d *Docker) Gather(acc telegraf.Accumulator) error {
 		}
 	}
 
+	filterArgs := filters.NewArgs()
+	for _, state := range containerStates {
+		if d.stateFilter.Match(state) {
+			filterArgs.Add("status", state)
+		}
+	}
+
+	// All container states were excluded
+	if filterArgs.Len() == 0 {
+		return nil
+	}
+
 	// List containers
-	opts := types.ContainerListOptions{}
+	opts := types.ContainerListOptions{
+		Filters: filterArgs,
+	}
 	ctx, cancel := context.WithTimeout(context.Background(), d.Timeout.Duration)
 	defer cancel()
 	containers, err := d.client.ContainerList(ctx, opts)
@@ -768,6 +797,18 @@ func (d *Docker) createLabelFilters() error {
 	return nil
 }
 
+func (d *Docker) createContainerStateFilters() error {
+	if len(d.ContainerStateInclude) == 0 && len(d.ContainerStateExclude) == 0 {
+		d.ContainerStateInclude = []string{"running"}
+	}
+	filter, err := filter.NewIncludeExcludeFilter(d.ContainerStateInclude, d.ContainerStateExclude)
+	if err != nil {
+		return err
+	}
+	d.stateFilter = filter
+	return nil
+}
+
 func init() {
 	inputs.Add("docker", func() telegraf.Input {
 		return &Docker{

plugins/inputs/docker/docker_test.go

@@ -3,6 +3,7 @@ package docker
 import (
 	"context"
 	"crypto/tls"
+	"sort"
 	"testing"
 
 	"github.com/influxdata/telegraf/testutil"
@@ -711,3 +712,85 @@ func TestDockerGatherSwarmInfo(t *testing.T) {
 		},
 	)
 }
+
+func TestContainerStateFilter(t *testing.T) {
+	var tests = []struct {
+		name     string
+		include  []string
+		exclude  []string
+		expected map[string][]string
+	}{
+		{
+			name: "default",
+			expected: map[string][]string{
+				"status": []string{"running"},
+			},
+		},
+		{
+			name:    "include running",
+			include: []string{"running"},
+			expected: map[string][]string{
+				"status": []string{"running"},
+			},
+		},
+		{
+			name:    "include glob",
+			include: []string{"r*"},
+			expected: map[string][]string{
+				"status": []string{"restarting", "running", "removing"},
+			},
+		},
+		{
+			name:    "include all",
+			include: []string{"*"},
+			expected: map[string][]string{
+				"status": []string{"created", "restarting", "running", "removing", "paused", "exited", "dead"},
+			},
+		},
+		{
+			name:    "exclude all",
+			exclude: []string{"*"},
+			expected: map[string][]string{
+				"status": []string{},
+			},
+		},
+		{
+			name:    "exclude all",
+			include: []string{"*"},
+			exclude: []string{"exited"},
+			expected: map[string][]string{
+				"status": []string{"created", "restarting", "running", "removing", "paused", "dead"},
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			var acc testutil.Accumulator
+
+			newClientFunc := func(host string, tlsConfig *tls.Config) (Client, error) {
+				client := baseClient
+				client.ContainerListF = func(ctx context.Context, options types.ContainerListOptions) ([]types.Container, error) {
+					for k, v := range tt.expected {
+						actual := options.Filters.Get(k)
+						sort.Strings(actual)
+						sort.Strings(v)
+						require.Equal(t, v, actual)
+					}
+
+					return nil, nil
+				}
+				return &client, nil
+			}
+
+			d := Docker{
+				newClient:             newClientFunc,
+				ContainerStateInclude: tt.include,
+				ContainerStateExclude: tt.exclude,
+			}
+
+			err := d.Gather(&acc)
+			require.NoError(t, err)
+		})
+	}
+}