Merge pull request #848 from linxiulei/revert

Support revert-pattern in logcounter

cmd/logcounter/options/options.go

@@ -34,6 +34,7 @@ type LogCounterOptions struct {
 	Lookback       string
 	Delay          string
 	Pattern        string
+	RevertPattern  string
 	Count          int
 }
 
@@ -46,6 +47,8 @@ func (fedo *LogCounterOptions) AddFlags(fs *pflag.FlagSet) {
 		"The time duration log watcher delays after node boot time. This is useful when log watcher needs to wait for some time until the node is stable.")
 	fs.StringVar(&fedo.Pattern, "pattern", "",
 		"The regular expression to match the problem in log. The pattern must match to the end of the line.")
+	fs.StringVar(&fedo.RevertPattern, "revert-pattern", "",
+		"Similar to --pattern but conversely it decreases count value for every match. This is useful to discount a log when another log occurs.")
 	fs.IntVar(&fedo.Count, "count", 1,
 		"The number of times the pattern must be found to trigger the condition")
 }

config/systemd-monitor-counter.json

@@ -37,7 +37,8 @@
         "--lookback=20m",
         "--delay=5m",
         "--count=5",
-        "--pattern=Started (Kubernetes kubelet|kubelet.service|kubelet.service - Kubernetes kubelet)."
+        "--pattern=Started (Kubernetes kubelet|kubelet.service|kubelet.service - Kubernetes kubelet).",
+        "--revert-pattern=Stopping (Kubernetes kubelet|kubelet.service|kubelet.service - Kubernetes kubelet)..."
       ],
       "timeout": "1m"
     },
@@ -51,7 +52,8 @@
         "--log-path=/var/log/journal",
         "--lookback=20m",
         "--count=5",
-        "--pattern=Starting (Docker Application Container Engine|docker.service|docker.service - Docker Application Container Engine)..."
+        "--pattern=Starting (Docker Application Container Engine|docker.service|docker.service - Docker Application Container Engine)...",
+        "--revert-pattern=Stopping (Docker Application Container Engine|docker.service|docker.service - Docker Application Container Engine)..."
       ],
       "timeout": "1m"
     },
@@ -65,7 +67,8 @@
         "--log-path=/var/log/journal",
         "--lookback=20m",
         "--count=5",
-        "--pattern=Starting (containerd container runtime|containerd.service|containerd.service - containerd container runtime)..."
+        "--pattern=Starting (containerd container runtime|containerd.service|containerd.service - containerd container runtime)...",
+        "--revert-pattern=Stopping (containerd container runtime|containerd.service|containerd.service - containerd container runtime)..."
       ],
       "timeout": "1m"
     }

pkg/logcounter/log_counter.go

@@ -40,10 +40,11 @@ const (
 )
 
 type logCounter struct {
-	logCh   <-chan *systemtypes.Log
-	buffer  systemlogmonitor.LogBuffer
-	pattern string
-	clock   clock.Clock
+	logCh         <-chan *systemtypes.Log
+	buffer        systemlogmonitor.LogBuffer
+	pattern       string
+	revertPattern string
+	clock         clock.Clock
 }
 
 func NewJournaldLogCounter(options *options.LogCounterOptions) (types.LogCounter, error) {
@@ -59,10 +60,11 @@ func NewJournaldLogCounter(options *options.LogCounterOptions) (types.LogCounter
 		return nil, fmt.Errorf("error watching journald: %v", err)
 	}
 	return &logCounter{
-		logCh:   logCh,
-		buffer:  systemlogmonitor.NewLogBuffer(bufferSize),
-		pattern: options.Pattern,
-		clock:   clock.RealClock{},
+		logCh:         logCh,
+		buffer:        systemlogmonitor.NewLogBuffer(bufferSize),
+		pattern:       options.Pattern,
+		revertPattern: options.RevertPattern,
+		clock:         clock.RealClock{},
 	}, nil
 }
 
@@ -84,6 +86,9 @@ func (e *logCounter) Count() (count int, err error) {
 			if len(e.buffer.Match(e.pattern)) != 0 {
 				count++
 			}
+			if e.revertPattern != "" && len(e.buffer.Match(e.revertPattern)) != 0 {
+				count--
+			}
 		case <-e.clock.After(timeout):
 			// Don't block forever if we do not get any new messages
 			return