Merge metrics descriptions in textfile collector (prometheus#2475)

The textfile collector will now provide a unified metric description
(that will look like "Metric read from file/a.prom, file/b.prom")
for metrics collected accross several text-files that don't already
have a description.

Also change the error handling in the textfile collector tests to
ContinueOnError to better mirror the real-life use-case.

Signed-off-by: Guillaume Espanel <guillaume.espanel.ext@ovhcloud.com>

Signed-off-by: Guillaume Espanel <guillaume.espanel.ext@ovhcloud.com>

collector/fixtures/textfile/metrics_merge_different_help.out

@@ -0,0 +1,11 @@
+# HELP events_total A nice help message.
+# TYPE events_total counter
+events_total{file="a",foo="bar"} 10
+events_total{file="a",foo="baz"} 20
+# HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
+# TYPE node_textfile_mtime_seconds gauge
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_different_help/a.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_different_help/b.prom"} 1
+# HELP node_textfile_scrape_error 1 if there was an error opening or reading a file, 0 otherwise
+# TYPE node_textfile_scrape_error gauge
+node_textfile_scrape_error 0

collector/fixtures/textfile/metrics_merge_different_help/a.prom

@@ -0,0 +1,5 @@
+# HELP events_total A nice help message.
+# TYPE events_total counter
+events_total{foo="bar",file="a"} 10
+events_total{foo="baz",file="a"} 20
+

collector/fixtures/textfile/metrics_merge_different_help/b.prom

@@ -0,0 +1,5 @@
+# HELP events_total A different help message.
+# TYPE events_total counter
+events_total{foo="bar",file="b"} 30
+events_total{foo="baz",file="b"} 40
+

collector/fixtures/textfile/metrics_merge_empty_help.out

@@ -0,0 +1,13 @@
+# HELP events_total Metric read from fixtures/textfile/metrics_merge_empty_help/a.prom, fixtures/textfile/metrics_merge_empty_help/b.prom
+# TYPE events_total counter
+events_total{file="a",foo="bar"} 10
+events_total{file="a",foo="baz"} 20
+events_total{file="b",foo="bar"} 30
+events_total{file="b",foo="baz"} 40
+# HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
+# TYPE node_textfile_mtime_seconds gauge
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_empty_help/a.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_empty_help/b.prom"} 1
+# HELP node_textfile_scrape_error 1 if there was an error opening or reading a file, 0 otherwise
+# TYPE node_textfile_scrape_error gauge
+node_textfile_scrape_error 0

collector/fixtures/textfile/metrics_merge_empty_help/a.prom

@@ -0,0 +1,5 @@
+# HELP events_total 
+# TYPE events_total counter
+events_total{foo="bar",file="a"} 10
+events_total{foo="baz",file="a"} 20
+

collector/fixtures/textfile/metrics_merge_empty_help/b.prom

@@ -0,0 +1,5 @@
+# HELP events_total 
+# TYPE events_total counter
+events_total{foo="bar",file="b"} 30
+events_total{foo="baz",file="b"} 40
+

collector/fixtures/textfile/metrics_merge_no_help.out

@@ -0,0 +1,13 @@
+# HELP events_total Metric read from fixtures/textfile/metrics_merge_no_help/a.prom, fixtures/textfile/metrics_merge_no_help/b.prom
+# TYPE events_total counter
+events_total{file="a",foo="bar"} 10
+events_total{file="a",foo="baz"} 20
+events_total{file="b",foo="bar"} 30
+events_total{file="b",foo="baz"} 40
+# HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
+# TYPE node_textfile_mtime_seconds gauge
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_no_help/a.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_no_help/b.prom"} 1
+# HELP node_textfile_scrape_error 1 if there was an error opening or reading a file, 0 otherwise
+# TYPE node_textfile_scrape_error gauge
+node_textfile_scrape_error 0

collector/fixtures/textfile/metrics_merge_no_help/a.prom

@@ -0,0 +1,4 @@
+# TYPE events_total counter
+events_total{foo="bar",file="a"} 10
+events_total{foo="baz",file="a"} 20
+

collector/fixtures/textfile/metrics_merge_no_help/b.prom

@@ -0,0 +1,4 @@
+# TYPE events_total counter
+events_total{foo="bar",file="b"} 30
+events_total{foo="baz",file="b"} 40
+

collector/fixtures/textfile/metrics_merge_same_help.out

@@ -0,0 +1,13 @@
+# HELP events_total The same help.
+# TYPE events_total counter
+events_total{file="a",foo="bar"} 10
+events_total{file="a",foo="baz"} 20
+events_total{file="b",foo="bar"} 30
+events_total{file="b",foo="baz"} 40
+# HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
+# TYPE node_textfile_mtime_seconds gauge
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_same_help/a.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/metrics_merge_same_help/b.prom"} 1
+# HELP node_textfile_scrape_error 1 if there was an error opening or reading a file, 0 otherwise
+# TYPE node_textfile_scrape_error gauge
+node_textfile_scrape_error 0

collector/fixtures/textfile/metrics_merge_same_help/a.prom

@@ -0,0 +1,5 @@
+# HELP events_total The same help.
+# TYPE events_total counter
+events_total{foo="bar",file="a"} 10
+events_total{foo="baz",file="a"} 20
+

collector/fixtures/textfile/metrics_merge_same_help/b.prom

@@ -0,0 +1,5 @@
+# HELP events_total The same help.
+# TYPE events_total counter
+events_total{foo="bar",file="b"} 30
+events_total{foo="baz",file="b"} 40
+

collector/textfile.go

@@ -191,6 +191,8 @@ func (c *textFileCollector) Update(ch chan<- prometheus.Metric) error {
 	// Iterate over files and accumulate their metrics, but also track any
 	// parsing errors so an error metric can be reported.
 	var errored bool
+	var parsedFamilies []*dto.MetricFamily
+	metricsNamesToFiles := map[string][]string{}
 
 	paths, err := filepath.Glob(c.path)
 	if err != nil || len(paths) == 0 {
@@ -208,20 +210,39 @@ func (c *textFileCollector) Update(ch chan<- prometheus.Metric) error {
 		}
 
 		for _, f := range files {
+			metricsFilePath := filepath.Join(path, f.Name())
 			if !strings.HasSuffix(f.Name(), ".prom") {
 				continue
 			}
 
-			mtime, err := c.processFile(path, f.Name(), ch)
+			mtime, families, err := c.processFile(path, f.Name(), ch)
+
+			for _, mf := range families {
+				metricsNamesToFiles[*mf.Name] = append(metricsNamesToFiles[*mf.Name], metricsFilePath)
+				parsedFamilies = append(parsedFamilies, mf)
+			}
+
 			if err != nil {
 				errored = true
 				level.Error(c.logger).Log("msg", "failed to collect textfile data", "file", f.Name(), "err", err)
 				continue
 			}
 
-			mtimes[filepath.Join(path, f.Name())] = *mtime
+			mtimes[metricsFilePath] = *mtime
 		}
 	}
+
+	for _, mf := range parsedFamilies {
+		if mf.Help == nil {
+			help := fmt.Sprintf("Metric read from %s", strings.Join(metricsNamesToFiles[*mf.Name], ", "))
+			mf.Help = &help
+		}
+	}
+
+	for _, mf := range parsedFamilies {
+		convertMetricFamily(mf, ch, c.logger)
+	}
+
 	c.exportMTimes(mtimes, ch)
 
 	// Export if there were errors.
@@ -243,44 +264,33 @@ func (c *textFileCollector) Update(ch chan<- prometheus.Metric) error {
 }
 
 // processFile processes a single file, returning its modification time on success.
-func (c *textFileCollector) processFile(dir, name string, ch chan<- prometheus.Metric) (*time.Time, error) {
+func (c *textFileCollector) processFile(dir, name string, ch chan<- prometheus.Metric) (*time.Time, map[string]*dto.MetricFamily, error) {
 	path := filepath.Join(dir, name)
 	f, err := os.Open(path)
 	if err != nil {
-		return nil, fmt.Errorf("failed to open textfile data file %q: %w", path, err)
+		return nil, nil, fmt.Errorf("failed to open textfile data file %q: %w", path, err)
 	}
 	defer f.Close()
 
 	var parser expfmt.TextParser
 	families, err := parser.TextToMetricFamilies(f)
 	if err != nil {
-		return nil, fmt.Errorf("failed to parse textfile data from %q: %w", path, err)
+		return nil, nil, fmt.Errorf("failed to parse textfile data from %q: %w", path, err)
 	}
 
 	if hasTimestamps(families) {
-		return nil, fmt.Errorf("textfile %q contains unsupported client-side timestamps, skipping entire file", path)
-	}
-
-	for _, mf := range families {
-		if mf.Help == nil {
-			help := fmt.Sprintf("Metric read from %s", path)
-			mf.Help = &help
-		}
-	}
-
-	for _, mf := range families {
-		convertMetricFamily(mf, ch, c.logger)
+		return nil, nil, fmt.Errorf("textfile %q contains unsupported client-side timestamps, skipping entire file", path)
 	}
 
 	// Only stat the file once it has been parsed and validated, so that
 	// a failure does not appear fresh.
 	stat, err := f.Stat()
 	if err != nil {
-		return nil, fmt.Errorf("failed to stat %q: %w", path, err)
+		return nil, families, fmt.Errorf("failed to stat %q: %w", path, err)
 	}
 
 	t := stat.ModTime()
-	return &t, nil
+	return &t, families, nil
 }
 
 // hasTimestamps returns true when metrics contain unsupported timestamps.

collector/textfile_test.go

@@ -95,6 +95,22 @@ func TestTextfileCollector(t *testing.T) {
 			path: "fixtures/textfile/*_extra_dimension",
 			out:  "fixtures/textfile/glob_extra_dimension.out",
 		},
+		{
+			path: "fixtures/textfile/metrics_merge_empty_help",
+			out:  "fixtures/textfile/metrics_merge_empty_help.out",
+		},
+		{
+			path: "fixtures/textfile/metrics_merge_no_help",
+			out:  "fixtures/textfile/metrics_merge_no_help.out",
+		},
+		{
+			path: "fixtures/textfile/metrics_merge_same_help",
+			out:  "fixtures/textfile/metrics_merge_same_help.out",
+		},
+		{
+			path: "fixtures/textfile/metrics_merge_different_help",
+			out:  "fixtures/textfile/metrics_merge_different_help.out",
+		},
 	}
 
 	for i, test := range tests {
@@ -117,7 +133,7 @@ func TestTextfileCollector(t *testing.T) {
 		registry.MustRegister(collectorAdapter{c})
 
 		rw := httptest.NewRecorder()
-		promhttp.HandlerFor(registry, promhttp.HandlerOpts{}).ServeHTTP(rw, &http.Request{})
+		promhttp.HandlerFor(registry, promhttp.HandlerOpts{ErrorHandling: promhttp.ContinueOnError}).ServeHTTP(rw, &http.Request{})
 		got := string(rw.Body.String())
 
 		want, err := os.ReadFile(test.out)