diff --git a/CHANGELOG.asciidoc b/CHANGELOG.asciidoc
index 4b34f1010b76..c2826be5b452 100644
--- a/CHANGELOG.asciidoc
+++ b/CHANGELOG.asciidoc
@@ -38,6 +38,7 @@ https://github.com/elastic/beats/compare/v5.0.0-beta1...master[Check the HEAD di
 *Topbeat*
 
 *Filebeat*
+- Fix input buffer on encoding problem
 
 *Winlogbeat*
 
diff --git a/filebeat/harvester/reader/line.go b/filebeat/harvester/reader/line.go
index ca7aac09d57b..de9c0e62e4e3 100644
--- a/filebeat/harvester/reader/line.go
+++ b/filebeat/harvester/reader/line.go
@@ -131,6 +131,8 @@ func (l *Line) advance() error {
 	sz, err := l.decode(idx + len(l.nl))
 	if err != nil {
 		logp.Err("Error decoding line: %s", err)
+		// In case of error increase size by unencoded length
+		sz = idx + len(l.nl)
 	}
 
 	// consume transformed bytes from input buffer
@@ -157,19 +159,20 @@ func (l *Line) decode(end int) (int, error) {
 		var nDst, nSrc int
 
 		nDst, nSrc, err = l.decoder.Transform(buffer, inBytes[start:end], false)
-
-		start += nSrc
-
-		l.outBuffer.Write(buffer[:nDst])
-
 		if err != nil {
-			if err == transform.ErrShortDst { // continue transforming
-				// Reset error as decoding continues
-				err = nil
-				continue
+			// Check if error is different from destination buffer too short
+			if err != transform.ErrShortDst {
+				l.outBuffer.Write(inBytes[0:end])
+				start = end
+				break
 			}
-			break
+
+			// Reset error as decoding continues
+			err = nil
 		}
+
+		start += nSrc
+		l.outBuffer.Write(buffer[:nDst])
 	}
 
 	l.byteCount += start
diff --git a/filebeat/tests/system/test_harvester.py b/filebeat/tests/system/test_harvester.py
index 9af6c1119a98..fa4fcef612f0 100644
--- a/filebeat/tests/system/test_harvester.py
+++ b/filebeat/tests/system/test_harvester.py
@@ -1,3 +1,5 @@
+# coding=utf-8
+
 from filebeat import BaseTest
 import os
 import codecs
@@ -756,3 +758,48 @@ def test_truncate(self):
         # Check that only 1 registry entry as original was only truncated
         data = self.get_registry()
         assert len(data) == 1
+
+
+    def test_decode_error(self):
+        """
+        Tests that in case of a decoding error it is handled gracefully
+        """
+        self.render_config_template(
+            path=os.path.abspath(self.working_dir) + "/log/*",
+            encoding="GBK", # Set invalid encoding for entry below which is actually uft-8
+        )
+
+        os.mkdir(self.working_dir + "/log/")
+
+        logfile = self.working_dir + "/log/test.log"
+
+        with open(logfile, 'w') as file:
+            file.write("hello world1" + "\n")
+
+            file.write('<meta content="瞭解「Google 商業解決方案」提供的各類服務軟件如何助您分析資料、刊登廣告、提升網站成效等。" name="description">' + '\n')
+            file.write("hello world2" + "\n")
+
+        filebeat = self.start_beat()
+
+        # Make sure both files were read
+        self.wait_until(
+            lambda: self.output_has(lines=3),
+            max_timeout=10)
+
+        # Wait until error shows up
+        self.wait_until(
+            lambda: self.log_contains("Error decoding line: simplifiedchinese: invalid GBK encoding"),
+            max_timeout=5)
+
+        filebeat.check_kill_and_wait()
+
+        # Check that only 1 registry entry as original was only truncated
+        data = self.get_registry()
+        assert len(data) == 1
+
+        output = self.read_output_json()
+        assert output[2]["message"] == "hello world2"
+
+
+
+