Add: merge by directory, Simplified: timeline merges

src/go/pt-galera-log-explainer/main.go

@@ -32,6 +32,7 @@ var CLI struct {
 	Verbosity        types.Verbosity `type:"counter" short:"v" default:"1" help:"-v: Detailed (default), -vv: DebugMySQL (add every mysql info the tool used), -vvv: Debug (internal tool debug)"`
 	PxcOperator      bool            `default:"false" help:"Analyze logs from Percona PXC operator. Off by default because it negatively impacts performance for non-k8s setups"`
 	ExcludeRegexes   []string        `help:"Remove regexes from analysis. List regexes using 'pt-galera-log-explainer regex-list'"`
+	MergeByDirectory bool            `help:"Instead of relying on identification, merge contexts and columns by base directory. Very useful when dealing with many small logs organized per directories."`
 
 	List      list       `cmd:""`
 	Whois     whois      `cmd:""`
@@ -92,29 +93,17 @@ func timelineFromPaths(paths []string, toCheck types.RegexMap, since, until *tim
 		found = true
 		extr.logger.Debug().Str("path", path).Msg("Finished searching")
 
-		// identify the node with the easiest to read information
-		// this is critical part to aggregate logs: this is what enable to merge logs
-		// ultimately the "identifier" will be used for columns header
-		var node string
-		if CLI.PxcOperator {
-			node = path
-
-		} else {
-
 		// Why it should not just identify using the file path:
 		// so that we are able to merge files that belong to the same nodes
 		// we wouldn't want them to be shown as from different nodes
-			node = types.Identifier(localTimeline[len(localTimeline)-1].Ctx)
+		if CLI.PxcOperator {
-			if t, ok := timeline[node]; ok {
+			timeline[path] = localTimeline
-
+		} else if CLI.MergeByDirectory {
-				extr.logger.Debug().Str("path", path).Str("node", node).Msg("Merging with existing timeline")
+			timeline.MergeByDirectory(path, localTimeline)
-				localTimeline = types.MergeTimeline(t, localTimeline)
+		} else {
+			timeline.MergeByIdentifier(localTimeline)
 		}
 	}
-		extr.logger.Debug().Str("path", path).Str("node", node).Msg("Storing timeline")
-		timeline[node] = localTimeline
-
-	}
 	if !found {
 		return nil, errors.New("Could not find data")
 	}

src/go/pt-galera-log-explainer/types/timeline.go

@@ -2,6 +2,7 @@ package types
 
 import (
 	"math"
+	"path/filepath"
 	"time"
 )
 
@@ -26,6 +27,27 @@ func (lt LocalTimeline) Add(li LogInfo) LocalTimeline {
 // "string" key is a node IP
 type Timeline map[string]LocalTimeline
 
+func (timeline Timeline) MergeByIdentifier(lt LocalTimeline) {
+	// identify the node with the easiest to read information
+	// this is critical part to aggregate logs: this is what enable to merge logs
+	// ultimately the "identifier" will be used for columns header
+	node := Identifier(lt[len(lt)-1].Ctx)
+	if lt2, ok := timeline[node]; ok {
+		lt = MergeTimeline(lt2, lt)
+	}
+	timeline[node] = lt
+}
+
+func (timeline Timeline) MergeByDirectory(path string, lt LocalTimeline) {
+	node := filepath.Base(filepath.Dir(path))
+	for _, lt2 := range timeline {
+		if len(lt2) > 0 && node == filepath.Base(filepath.Dir(lt2[0].Ctx.FilePath)) {
+			lt = MergeTimeline(lt2, lt)
+		}
+	}
+	timeline[node] = lt
+}
+
 // MergeTimeline is helpful when log files are split by date, it can be useful to be able to merge content
 // a "timeline" come from a log file. Log files that came from some node should not never have overlapping dates
 func MergeTimeline(t1, t2 LocalTimeline) LocalTimeline {