Merge pull request #12019 from roidelapluie/scrape-config-files

Add include scrape configs

cmd/prometheus/main.go

@@ -468,6 +468,14 @@ func main() {
 		level.Error(logger).Log("msg", fmt.Sprintf("Error loading config (--config.file=%s)", cfg.configFile), "file", absPath, "err", err)
 		os.Exit(2)
 	}
+	if _, err := cfgFile.GetScrapeConfigs(); err != nil {
+		absPath, pathErr := filepath.Abs(cfg.configFile)
+		if pathErr != nil {
+			absPath = cfg.configFile
+		}
+		level.Error(logger).Log("msg", fmt.Sprintf("Error loading scrape config files from config (--config.file=%q)", cfg.configFile), "file", absPath, "err", err)
+		os.Exit(2)
+	}
 	if cfg.tsdb.EnableExemplarStorage {
 		if cfgFile.StorageConfig.ExemplarsConfig == nil {
 			cfgFile.StorageConfig.ExemplarsConfig = &config.DefaultExemplarsConfig
@@ -730,7 +738,11 @@ func main() {
 			name: "scrape_sd",
 			reloader: func(cfg *config.Config) error {
 				c := make(map[string]discovery.Configs)
-				for _, v := range cfg.ScrapeConfigs {
+				scfgs, err := cfg.GetScrapeConfigs()
+				if err != nil {
+					return err
+				}
+				for _, v := range scfgs {
 					c[v.JobName] = v.ServiceDiscoveryConfigs
 				}
 				return discoveryManagerScrape.ApplyConfig(c)

cmd/promtool/main.go

@@ -463,7 +463,18 @@ func checkConfig(agentMode bool, filename string, checkSyntaxOnly bool) ([]strin
 		}
 	}
 
-	for _, scfg := range cfg.ScrapeConfigs {
+	var scfgs []*config.ScrapeConfig
+	if checkSyntaxOnly {
+		scfgs = cfg.ScrapeConfigs
+	} else {
+		var err error
+		scfgs, err = cfg.GetScrapeConfigs()
+		if err != nil {
+			return nil, fmt.Errorf("error loading scrape configs: %w", err)
+		}
+	}
+
+	for _, scfg := range scfgs {
 		if !checkSyntaxOnly && scfg.HTTPClientConfig.Authorization != nil {
 			if err := checkFileExists(scfg.HTTPClientConfig.Authorization.CredentialsFile); err != nil {
 				return nil, fmt.Errorf("error checking authorization credentials or bearer token file %q: %w", scfg.HTTPClientConfig.Authorization.CredentialsFile, err)

cmd/promtool/sd.go

@@ -47,9 +47,15 @@ func CheckSD(sdConfigFiles, sdJobName string, sdTimeout time.Duration, noDefault
 	}
 
 	var scrapeConfig *config.ScrapeConfig
+	scfgs, err := cfg.GetScrapeConfigs()
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "Cannot load scrape configs", err)
+		return failureExitCode
+	}
+
 	jobs := []string{}
 	jobMatched := false
-	for _, v := range cfg.ScrapeConfigs {
+	for _, v := range scfgs {
 		jobs = append(jobs, v.JobName)
 		if v.JobName == sdJobName {
 			jobMatched = true

config/config.go

@@ -216,12 +216,13 @@ var (
 
 // Config is the top-level configuration for Prometheus's config files.
 type Config struct {
-	GlobalConfig   GlobalConfig    `yaml:"global"`
+	GlobalConfig      GlobalConfig    `yaml:"global"`
-	AlertingConfig AlertingConfig  `yaml:"alerting,omitempty"`
+	AlertingConfig    AlertingConfig  `yaml:"alerting,omitempty"`
-	RuleFiles      []string        `yaml:"rule_files,omitempty"`
+	RuleFiles         []string        `yaml:"rule_files,omitempty"`
-	ScrapeConfigs  []*ScrapeConfig `yaml:"scrape_configs,omitempty"`
+	ScrapeConfigFiles []string        `yaml:"scrape_config_files,omitempty"`
-	StorageConfig  StorageConfig   `yaml:"storage,omitempty"`
+	ScrapeConfigs     []*ScrapeConfig `yaml:"scrape_configs,omitempty"`
-	TracingConfig  TracingConfig   `yaml:"tracing,omitempty"`
+	StorageConfig     StorageConfig   `yaml:"storage,omitempty"`
+	TracingConfig     TracingConfig   `yaml:"tracing,omitempty"`
 
 	RemoteWriteConfigs []*RemoteWriteConfig `yaml:"remote_write,omitempty"`
 	RemoteReadConfigs  []*RemoteReadConfig  `yaml:"remote_read,omitempty"`
@@ -235,6 +236,9 @@ func (c *Config) SetDirectory(dir string) {
 	for i, file := range c.RuleFiles {
 		c.RuleFiles[i] = config.JoinDir(dir, file)
 	}
+	for i, file := range c.ScrapeConfigFiles {
+		c.ScrapeConfigFiles[i] = config.JoinDir(dir, file)
+	}
 	for _, c := range c.ScrapeConfigs {
 		c.SetDirectory(dir)
 	}
@@ -254,6 +258,58 @@ func (c Config) String() string {
 	return string(b)
 }
 
+// ScrapeConfigs returns the scrape configurations.
+func (c *Config) GetScrapeConfigs() ([]*ScrapeConfig, error) {
+	scfgs := make([]*ScrapeConfig, len(c.ScrapeConfigs))
+
+	jobNames := map[string]string{}
+	for i, scfg := range c.ScrapeConfigs {
+		// We do these checks for library users that would not call Validate in
+		// Unmarshal.
+		if err := scfg.Validate(c.GlobalConfig.ScrapeInterval, c.GlobalConfig.ScrapeTimeout); err != nil {
+			return nil, err
+		}
+
+		if _, ok := jobNames[scfg.JobName]; ok {
+			return nil, fmt.Errorf("found multiple scrape configs with job name %q", scfg.JobName)
+		}
+		jobNames[scfg.JobName] = "main config file"
+		scfgs[i] = scfg
+	}
+	for _, pat := range c.ScrapeConfigFiles {
+		fs, err := filepath.Glob(pat)
+		if err != nil {
+			// The only error can be a bad pattern.
+			return nil, fmt.Errorf("error retrieving scrape config files for %q: %w", pat, err)
+		}
+		for _, filename := range fs {
+			cfg := ScrapeConfigs{}
+			content, err := os.ReadFile(filename)
+			if err != nil {
+				return nil, fileErr(filename, err)
+			}
+			err = yaml.UnmarshalStrict(content, &cfg)
+			if err != nil {
+				return nil, fileErr(filename, err)
+			}
+			for _, scfg := range cfg.ScrapeConfigs {
+				if err := scfg.Validate(c.GlobalConfig.ScrapeInterval, c.GlobalConfig.ScrapeTimeout); err != nil {
+					return nil, fileErr(filename, err)
+				}
+
+				if f, ok := jobNames[scfg.JobName]; ok {
+					return nil, fileErr(filename, fmt.Errorf("found multiple scrape configs with job name %q, first found in %s", scfg.JobName, f))
+				}
+				jobNames[scfg.JobName] = fmt.Sprintf("%q", filePath(filename))
+
+				scfg.SetDirectory(filepath.Dir(filename))
+				scfgs = append(scfgs, scfg)
+			}
+		}
+	}
+	return scfgs, nil
+}
+
 // UnmarshalYAML implements the yaml.Unmarshaler interface.
 func (c *Config) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	*c = DefaultConfig
@@ -276,26 +332,18 @@ func (c *Config) UnmarshalYAML(unmarshal func(interface{}) error) error {
 			return fmt.Errorf("invalid rule file path %q", rf)
 		}
 	}
+
+	for _, sf := range c.ScrapeConfigFiles {
+		if !patRulePath.MatchString(sf) {
+			return fmt.Errorf("invalid scrape config file path %q", sf)
+		}
+	}
+
 	// Do global overrides and validate unique names.
 	jobNames := map[string]struct{}{}
 	for _, scfg := range c.ScrapeConfigs {
-		if scfg == nil {
+		if err := scfg.Validate(c.GlobalConfig.ScrapeInterval, c.GlobalConfig.ScrapeTimeout); err != nil {
-			return errors.New("empty or null scrape config section")
+			return err
-		}
-		// First set the correct scrape interval, then check that the timeout
-		// (inferred or explicit) is not greater than that.
-		if scfg.ScrapeInterval == 0 {
-			scfg.ScrapeInterval = c.GlobalConfig.ScrapeInterval
-		}
-		if scfg.ScrapeTimeout > scfg.ScrapeInterval {
-			return fmt.Errorf("scrape timeout greater than scrape interval for scrape config with job name %q", scfg.JobName)
-		}
-		if scfg.ScrapeTimeout == 0 {
-			if c.GlobalConfig.ScrapeTimeout > scfg.ScrapeInterval {
-				scfg.ScrapeTimeout = scfg.ScrapeInterval
-			} else {
-				scfg.ScrapeTimeout = c.GlobalConfig.ScrapeTimeout
-			}
 		}
 
 		if _, ok := jobNames[scfg.JobName]; ok {
@@ -401,6 +449,10 @@ func (c *GlobalConfig) isZero() bool {
 		c.QueryLogFile == ""
 }
 
+type ScrapeConfigs struct {
+	ScrapeConfigs []*ScrapeConfig `yaml:"scrape_configs,omitempty"`
+}
+
 // ScrapeConfig configures a scraping unit for Prometheus.
 type ScrapeConfig struct {
 	// The job name to which the job label is set by default.
@@ -494,6 +546,28 @@ func (c *ScrapeConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	return nil
 }
 
+func (c *ScrapeConfig) Validate(defaultInterval, defaultTimeout model.Duration) error {
+	if c == nil {
+		return errors.New("empty or null scrape config section")
+	}
+	// First set the correct scrape interval, then check that the timeout
+	// (inferred or explicit) is not greater than that.
+	if c.ScrapeInterval == 0 {
+		c.ScrapeInterval = defaultInterval
+	}
+	if c.ScrapeTimeout > c.ScrapeInterval {
+		return fmt.Errorf("scrape timeout greater than scrape interval for scrape config with job name %q", c.JobName)
+	}
+	if c.ScrapeTimeout == 0 {
+		if defaultTimeout > c.ScrapeInterval {
+			c.ScrapeTimeout = c.ScrapeInterval
+		} else {
+			c.ScrapeTimeout = defaultTimeout
+		}
+	}
+	return nil
+}
+
 // MarshalYAML implements the yaml.Marshaler interface.
 func (c *ScrapeConfig) MarshalYAML() (interface{}, error) {
 	return discovery.MarshalYAMLWithInlineConfigs(c)
@@ -936,3 +1010,15 @@ func (c *RemoteReadConfig) UnmarshalYAML(unmarshal func(interface{}) error) erro
 	// Thus we just do its validation here.
 	return c.HTTPClientConfig.Validate()
 }
+
+func filePath(filename string) string {
+	absPath, err := filepath.Abs(filename)
+	if err != nil {
+		return filename
+	}
+	return absPath
+}
+
+func fileErr(filename string, err error) error {
+	return fmt.Errorf("%q: %w", filePath(filename), err)
+}

config/config_test.go

@@ -1693,6 +1693,10 @@ var expectedErrors = []struct {
 		filename: "ovhcloud_bad_service.bad.yml",
 		errMsg:   "unknown service: fakeservice",
 	},
+	{
+		filename: "scrape_config_files_glob.bad.yml",
+		errMsg:   `parsing YAML file testdata/scrape_config_files_glob.bad.yml: invalid scrape config file path "scrape_configs/*/*"`,
+	},
 }
 
 func TestBadConfigs(t *testing.T) {
@@ -1779,6 +1783,156 @@ func TestEmptyGlobalBlock(t *testing.T) {
 	require.Equal(t, exp, *c)
 }
 
+func TestGetScrapeConfigs(t *testing.T) {
+	sc := func(jobName string, scrapeInterval, scrapeTimeout model.Duration) *ScrapeConfig {
+		return &ScrapeConfig{
+			JobName:          jobName,
+			HonorTimestamps:  true,
+			ScrapeInterval:   scrapeInterval,
+			ScrapeTimeout:    scrapeTimeout,
+			MetricsPath:      "/metrics",
+			Scheme:           "http",
+			HTTPClientConfig: config.DefaultHTTPClientConfig,
+			ServiceDiscoveryConfigs: discovery.Configs{
+				discovery.StaticConfig{
+					{
+						Targets: []model.LabelSet{
+							{
+								model.AddressLabel: "localhost:8080",
+							},
+						},
+						Source: "0",
+					},
+				},
+			},
+		}
+	}
+
+	testCases := []struct {
+		name           string
+		configFile     string
+		expectedResult []*ScrapeConfig
+		expectedError  string
+	}{
+		{
+			name:           "An included config file should be a valid global config.",
+			configFile:     "testdata/scrape_config_files.good.yml",
+			expectedResult: []*ScrapeConfig{sc("prometheus", model.Duration(60*time.Second), model.Duration(10*time.Second))},
+		},
+		{
+			name:           "An global config that only include a scrape config file.",
+			configFile:     "testdata/scrape_config_files_only.good.yml",
+			expectedResult: []*ScrapeConfig{sc("prometheus", model.Duration(60*time.Second), model.Duration(10*time.Second))},
+		},
+		{
+			name:       "An global config that combine scrape config files and scrape configs.",
+			configFile: "testdata/scrape_config_files_combined.good.yml",
+			expectedResult: []*ScrapeConfig{
+				sc("node", model.Duration(60*time.Second), model.Duration(10*time.Second)),
+				sc("prometheus", model.Duration(60*time.Second), model.Duration(10*time.Second)),
+				sc("alertmanager", model.Duration(60*time.Second), model.Duration(10*time.Second)),
+			},
+		},
+		{
+			name:       "An global config that includes a scrape config file with globs",
+			configFile: "testdata/scrape_config_files_glob.good.yml",
+			expectedResult: []*ScrapeConfig{
+				{
+					JobName: "prometheus",
+
+					HonorTimestamps: true,
+					ScrapeInterval:  model.Duration(60 * time.Second),
+					ScrapeTimeout:   DefaultGlobalConfig.ScrapeTimeout,
+
+					MetricsPath: DefaultScrapeConfig.MetricsPath,
+					Scheme:      DefaultScrapeConfig.Scheme,
+
+					HTTPClientConfig: config.HTTPClientConfig{
+						TLSConfig: config.TLSConfig{
+							CertFile: filepath.FromSlash("testdata/scrape_configs/valid_cert_file"),
+							KeyFile:  filepath.FromSlash("testdata/scrape_configs/valid_key_file"),
+						},
+						FollowRedirects: true,
+						EnableHTTP2:     true,
+					},
+
+					ServiceDiscoveryConfigs: discovery.Configs{
+						discovery.StaticConfig{
+							{
+								Targets: []model.LabelSet{
+									{model.AddressLabel: "localhost:8080"},
+								},
+								Source: "0",
+							},
+						},
+					},
+				},
+				{
+					JobName: "node",
+
+					HonorTimestamps: true,
+					ScrapeInterval:  model.Duration(15 * time.Second),
+					ScrapeTimeout:   DefaultGlobalConfig.ScrapeTimeout,
+					HTTPClientConfig: config.HTTPClientConfig{
+						TLSConfig: config.TLSConfig{
+							CertFile: filepath.FromSlash("testdata/valid_cert_file"),
+							KeyFile:  filepath.FromSlash("testdata/valid_key_file"),
+						},
+						FollowRedirects: true,
+						EnableHTTP2:     true,
+					},
+
+					MetricsPath: DefaultScrapeConfig.MetricsPath,
+					Scheme:      DefaultScrapeConfig.Scheme,
+
+					ServiceDiscoveryConfigs: discovery.Configs{
+						&vultr.SDConfig{
+							HTTPClientConfig: config.HTTPClientConfig{
+								Authorization: &config.Authorization{
+									Type:        "Bearer",
+									Credentials: "abcdef",
+								},
+								FollowRedirects: true,
+								EnableHTTP2:     true,
+							},
+							Port:            80,
+							RefreshInterval: model.Duration(60 * time.Second),
+						},
+					},
+				},
+			},
+		},
+		{
+			name:          "An global config that includes twice the same scrape configs.",
+			configFile:    "testdata/scrape_config_files_double_import.bad.yml",
+			expectedError: `found multiple scrape configs with job name "prometheus"`,
+		},
+		{
+			name:          "An global config that includes a scrape config identical to a scrape config in the main file.",
+			configFile:    "testdata/scrape_config_files_duplicate.bad.yml",
+			expectedError: `found multiple scrape configs with job name "prometheus"`,
+		},
+		{
+			name:          "An global config that includes a scrape config file with errors.",
+			configFile:    "testdata/scrape_config_files_global.bad.yml",
+			expectedError: `scrape timeout greater than scrape interval for scrape config with job name "prometheus"`,
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			c, err := LoadFile(tc.configFile, false, false, log.NewNopLogger())
+			require.NoError(t, err)
+
+			scfgs, err := c.GetScrapeConfigs()
+			if len(tc.expectedError) > 0 {
+				require.ErrorContains(t, err, tc.expectedError)
+			}
+			require.Equal(t, tc.expectedResult, scfgs)
+		})
+	}
+}
+
 func kubernetesSDHostURL() config.URL {
 	tURL, _ := url.Parse("https://localhost:1234")
 	return config.URL{URL: tURL}

config/testdata/scrape_config_files.bad.yml

@@ -0,0 +1,6 @@
+scrape_configs:
+  - job_name: prometheus
+    scrape_interval: 10s
+    scrape_timeout: 20s
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files.good.yml

@@ -0,0 +1,4 @@
+scrape_configs:
+  - job_name: prometheus
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files2.good.yml

@@ -0,0 +1,4 @@
+scrape_configs:
+  - job_name: alertmanager
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files_combined.good.yml

@@ -0,0 +1,7 @@
+scrape_config_files:
+  - scrape_config_files.good.yml
+  - scrape_config_files2.good.yml
+scrape_configs:
+  - job_name: node
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files_double_import.bad.yml

@@ -0,0 +1,3 @@
+scrape_config_files:
+  - scrape_config_files.good.yml
+  - scrape_config_files.good.yml

config/testdata/scrape_config_files_duplicate.bad.yml

@@ -0,0 +1,6 @@
+scrape_config_files:
+  - scrape_config_files.good.yml
+scrape_configs:
+  - job_name: prometheus
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files_glob.bad.yml

@@ -0,0 +1,6 @@
+scrape_config_files:
+  - scrape_configs/*/*
+scrape_configs:
+  - job_name: node
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files_glob.good.yml

@@ -0,0 +1,2 @@
+scrape_config_files:
+  - scrape_configs/*.yml

config/testdata/scrape_config_files_global.bad.yml

@@ -0,0 +1,2 @@
+scrape_config_files:
+  - scrape_config_files.bad.yml

config/testdata/scrape_config_files_global_duplicate.bad.yml

@@ -0,0 +1,11 @@
+global:
+  scrape_interval: 15s
+
+scrape_config_files:
+  - scrape_config_files.good.yml
+  - scrape_config_files.good.yml
+
+scrape_configs:
+  - job_name: prometheus
+    static_configs:
+      - targets: ['localhost:8080']

config/testdata/scrape_config_files_only.good.yml

@@ -0,0 +1,2 @@
+scrape_config_files:
+  - scrape_config_files.good.yml

config/testdata/scrape_configs/scrape_config_files1.good.yml

@@ -0,0 +1,7 @@
+scrape_configs:
+  - job_name: prometheus
+    static_configs:
+      - targets: ['localhost:8080']
+    tls_config:
+      cert_file: valid_cert_file
+      key_file: valid_key_file

config/testdata/scrape_configs/scrape_config_files2.good.yml

@@ -0,0 +1,9 @@
+scrape_configs:
+  - job_name: node
+    scrape_interval: 15s
+    tls_config:
+      cert_file: ../valid_cert_file
+      key_file: ../valid_key_file
+    vultr_sd_configs:
+      - authorization:
+          credentials: abcdef

docs/configuration/configuration.md

@@ -78,6 +78,11 @@ global:
 rule_files:
   [ - <filepath_glob> ... ]
 
+# Scrape config files specifies a list of globs. Scrape configs are read from
+# all matching files and appended to the list of scrape configs.
+scrape_config_files:
+  [ - <filepath_glob> ... ]
+
 # A list of scrape configurations.
 scrape_configs:
   [ - <scrape_config> ... ]

scrape/manager.go

@@ -270,8 +270,13 @@ func (m *Manager) ApplyConfig(cfg *config.Config) error {
 	m.mtxScrape.Lock()
 	defer m.mtxScrape.Unlock()
 
+	scfgs, err := cfg.GetScrapeConfigs()
+	if err != nil {
+		return err
+	}
+
 	c := make(map[string]*config.ScrapeConfig)
-	for _, scfg := range cfg.ScrapeConfigs {
+	for _, scfg := range scfgs {
 		c[scfg.JobName] = scfg
 	}
 	m.scrapeConfigs = c