collector: Unwrap glob textfile directories (prometheus#1985)

* collector: Unwrap glob textfile directories
* collector: Store full path in mtime's file label

The point is to avoid duplicated gauges from files with the same name in
different directories.

This introduces support for exporting from multiple directories matching
given pattern (e.g. `/home/*/metrics/`).

Signed-off-by: Kiril Vladimirov <kiril@vladimiroff.org>

collector/fixtures/textfile/different_metric_types.out

@@ -26,7 +26,7 @@ events_total{foo="bar"} 10
 events_total{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="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/different_metric_types/metrics.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/glob_extra_dimension.out

@@ -0,0 +1,49 @@
+# HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
+# TYPE node_textfile_mtime_seconds gauge
+node_textfile_mtime_seconds{file="fixtures/textfile/histogram_extra_dimension/metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/summary_extra_dimension/metrics.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
+# HELP prometheus_rule_evaluation_duration_seconds The duration for a rule to execute.
+# TYPE prometheus_rule_evaluation_duration_seconds summary
+prometheus_rule_evaluation_duration_seconds{handler="",rule_type="alerting",quantile="0.9"} 0.001765451
+prometheus_rule_evaluation_duration_seconds{handler="",rule_type="alerting",quantile="0.99"} 0.018672076
+prometheus_rule_evaluation_duration_seconds_sum{handler="",rule_type="alerting"} 214.85081044700146
+prometheus_rule_evaluation_duration_seconds_count{handler="",rule_type="alerting"} 185209
+prometheus_rule_evaluation_duration_seconds{handler="",rule_type="recording",quantile="0.5"} 4.3132e-05
+prometheus_rule_evaluation_duration_seconds{handler="",rule_type="recording",quantile="0.9"} 8.9295e-05
+prometheus_rule_evaluation_duration_seconds{handler="",rule_type="recording",quantile="0.99"} 0.000193657
+prometheus_rule_evaluation_duration_seconds_sum{handler="",rule_type="recording"} 185091.01317759082
+prometheus_rule_evaluation_duration_seconds_count{handler="",rule_type="recording"} 1.0020195e+08
+prometheus_rule_evaluation_duration_seconds{handler="foo",rule_type="alerting",quantile="0.5"} 0.000571464
+prometheus_rule_evaluation_duration_seconds_sum{handler="foo",rule_type="alerting"} 0
+prometheus_rule_evaluation_duration_seconds_count{handler="foo",rule_type="alerting"} 0
+# HELP prometheus_tsdb_compaction_chunk_range Final time range of chunks on their first compaction
+# TYPE prometheus_tsdb_compaction_chunk_range histogram
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="100"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="400"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="1600"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="6400"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="25600"} 7
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="102400"} 7
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="409600"} 1.412839e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="1.6384e+06"} 1.69185e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="6.5536e+06"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="2.62144e+07"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="bar",le="+Inf"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_sum{foo="bar"} 6.71393432189e+11
+prometheus_tsdb_compaction_chunk_range_count{foo="bar"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="100"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="400"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="1600"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="6400"} 0
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="25600"} 7
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="102400"} 7
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="409600"} 1.412839e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="1.6384e+06"} 1.69185e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="6.5536e+06"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="2.62144e+07"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_bucket{foo="baz",le="+Inf"} 1.691853e+06
+prometheus_tsdb_compaction_chunk_range_sum{foo="baz"} 6.71393432189e+11
+prometheus_tsdb_compaction_chunk_range_count{foo="baz"} 1.691853e+06

collector/fixtures/textfile/histogram.out

@@ -1,6 +1,6 @@
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/histogram/metrics.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/histogram_extra_dimension.out

@@ -1,6 +1,6 @@
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/histogram_extra_dimension/metrics.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/inconsistent_metrics.out

@@ -23,7 +23,7 @@ http_requests_total{baz="",code="503",foo="",handler="query_range",method="get"}
 http_requests_total{baz="bar",code="200",foo="",handler="",method="get"} 93
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/inconsistent_metrics/metrics.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/summary.out

@@ -22,7 +22,7 @@ event_duration_seconds_total_sum{baz="result_sort"} 3.4123187829998307
 event_duration_seconds_total_count{baz="result_sort"} 1.427647e+06
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/summary/metrics.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/summary_extra_dimension.out

@@ -1,6 +1,6 @@
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/summary_extra_dimension/metrics.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/two_metric_files.out

@@ -1,7 +1,7 @@
 # HELP node_textfile_mtime_seconds Unixtime mtime of textfiles successfully read.
 # TYPE node_textfile_mtime_seconds gauge
-node_textfile_mtime_seconds{file="metrics1.prom"} 1
-node_textfile_mtime_seconds{file="metrics2.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/two_metric_files/metrics1.prom"} 1
+node_textfile_mtime_seconds{file="fixtures/textfile/two_metric_files/metrics2.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/textfile.go

@@ -172,18 +172,18 @@ func (c *textFileCollector) exportMTimes(mtimes map[string]time.Time, ch chan<-
 
 	// Export the mtimes of the successful files.
 	// Sorting is needed for predictable output comparison in tests.
-	filenames := make([]string, 0, len(mtimes))
-	for filename := range mtimes {
-		filenames = append(filenames, filename)
+	filepaths := make([]string, 0, len(mtimes))
+	for path := range mtimes {
+		filepaths = append(filepaths, path)
 	}
-	sort.Strings(filenames)
+	sort.Strings(filepaths)
 
-	for _, filename := range filenames {
-		mtime := float64(mtimes[filename].UnixNano() / 1e9)
+	for _, path := range filepaths {
+		mtime := float64(mtimes[path].UnixNano() / 1e9)
 		if c.mtime != nil {
 			mtime = *c.mtime
 		}
-		ch <- prometheus.MustNewConstMetric(mtimeDesc, prometheus.GaugeValue, mtime, filename)
+		ch <- prometheus.MustNewConstMetric(mtimeDesc, prometheus.GaugeValue, mtime, path)
 	}
 }
 
@@ -192,28 +192,37 @@ 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
-	files, err := ioutil.ReadDir(c.path)
-	if err != nil && c.path != "" {
-		errored = true
-		level.Error(c.logger).Log("msg", "failed to read textfile collector directory", "path", c.path, "err", err)
-	}
 
-	mtimes := make(map[string]time.Time, len(files))
-	for _, f := range files {
-		if !strings.HasSuffix(f.Name(), ".prom") {
-			continue
-		}
+	paths, err := filepath.Glob(c.path)
+	if err != nil || len(paths) == 0 {
+		// not glob or not accessible path either way assume single
+		// directory and let ioutil.ReadDir handle it
+		paths = []string{c.path}
+	}
 
-		mtime, err := c.processFile(f.Name(), ch)
-		if err != nil {
+	mtimes := make(map[string]time.Time)
+	for _, path := range paths {
+		files, err := ioutil.ReadDir(path)
+		if err != nil && path != "" {
 			errored = true
-			level.Error(c.logger).Log("msg", "failed to collect textfile data", "file", f.Name(), "err", err)
-			continue
+			level.Error(c.logger).Log("msg", "failed to read textfile collector directory", "path", path, "err", err)
 		}
 
-		mtimes[f.Name()] = *mtime
-	}
+		for _, f := range files {
+			if !strings.HasSuffix(f.Name(), ".prom") {
+				continue
+			}
 
+			mtime, err := c.processFile(path, f.Name(), ch)
+			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
+		}
+	}
 	c.exportMTimes(mtimes, ch)
 
 	// Export if there were errors.
@@ -235,8 +244,8 @@ func (c *textFileCollector) Update(ch chan<- prometheus.Metric) error {
 }
 
 // processFile processes a single file, returning its modification time on success.
-func (c *textFileCollector) processFile(name string, ch chan<- prometheus.Metric) (*time.Time, error) {
-	path := filepath.Join(c.path, name)
+func (c *textFileCollector) processFile(dir, name string, ch chan<- prometheus.Metric) (*time.Time, 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)

collector/textfile_test.go

@@ -91,6 +91,10 @@ func TestTextfileCollector(t *testing.T) {
 			path: "fixtures/textfile/summary_extra_dimension",
 			out:  "fixtures/textfile/summary_extra_dimension.out",
 		},
+		{
+			path: "fixtures/textfile/*_extra_dimension",
+			out:  "fixtures/textfile/glob_extra_dimension.out",
+		},
 	}
 
 	for i, test := range tests {