Log4j 2 Appender that ships logs using HTTPS bulk
This appender sends logs to your Logz.io account, using non-blocking threading, bulks, and HTTPS encryption. Please note that this appender requires log4j version 2.7 and up, and java 8 and up.
This appender uses LogzioSender implementation. Once you send a log, it will be enqueued in the queue and 100% non-blocking. There is a background task that will handle the log shipment for you. This jar is an "Uber-Jar" that shades both LogzioSender, BigQueue, Gson and Guava to avoid "dependency hell".
JDK 8:
<dependency>
<groupId>io.logz.log4j2</groupId>
<artifactId>logzio-log4j2-appender</artifactId>
<version>1.0.19</version>
</dependency>
JDK 11 and above:
<dependency>
<groupId>io.logz.log4j2</groupId>
<artifactId>logzio-log4j2-appender</artifactId>
<version>2.0.1</version>
</dependency>
The appender also requires a logger implementation, for example:
<dependency>
<groupId>org.apache.logging.log4j</groupId>
<artifactId>log4j-slf4j-impl</artifactId>
<version>2.15.0</version>
</dependency>
<Appenders>
<LogzioAppender name="Logzio">
<logzioToken>your-logzio-personal-token-from-settings</logzioToken>
<logzioType>myAwesomeType</logzioType>
<logzioUrl>https://listener.logz.io:8071</logzioUrl>
</LogzioAppender>
</Appenders>
<Loggers>
<Root level="info">
<AppenderRef ref="Logzio"/>
</Root>
</Loggers>
Parameter | Default | Explained |
---|---|---|
logzioToken | None | Your Logz.io token, which can be found under "settings" in your account, If the value begins with $ then the appender looks for an environment variable or system property with the name specified. For example: $LOGZIO_TOKEN will look for environment variable named LOGZIO_TOKEN |
logzioType | java | The log type for that appender, it must not contain any spaces |
logzioUrl | https://listener.logz.io:8071 | The url that the appender sends to. If your account is in the EU you must use https://listener-eu.logz.io:8071 |
drainTimeoutSec | 5 | How often the appender should drain the queue (in seconds) |
socketTimeoutMs | 10 * 1000 | The socket timeout during log shipment |
connectTimeoutMs | 10 * 1000 | The connection timeout during log shipment |
addHostname | false | Optional. If true, then a field named 'hostname' will be added holding the host name of the machine. If from some reason there's no defined hostname, this field won't be added |
additionalFields | None | Optional. Allows to add additional fields to the JSON message sent. The format is "fieldName1=fieldValue1;fieldName2=fieldValue2". You can optionally inject an environment variable value using the following format: "fieldName1=fieldValue1;fieldName2=$ENV_VAR_NAME". In that case, the environment variable should be the only value. In case the environment variable can't be resolved, the field will be omitted. |
debug | false | Print some debug messages to stdout to help to diagnose issues |
compressRequests | false | Boolean. true if logs are compressed in gzip format before sending. false if logs are sent uncompressed. |
exceedMaxSizeAction | "cut" | String. cut to truncate the message field or drop to drop log that exceed the allowed maximum size for logzio. If the log size exceeding the maximum size allowed after truncating the message field, the log will be dropped. |
Parameter | Default | Explained |
---|---|---|
inMemoryQueueCapacityBytes | 1024 * 1024 * 100 | The amount of memory(bytes) we are allowed to use for the memory queue. If the value is -1 the sender will not limit the queue size. |
inMemoryLogsCountCapacity | -1 | Number of logs we are allowed to have in the queue before dropping logs. If the value is -1 the sender will not limit the number of logs allowed. |
inMemoryQueue | false | Set to true if the appender uses in memory queue. By default the appender uses disk queue |
Parameter | Default | Explained |
---|---|---|
fileSystemFullPercentThreshold | 98 | The percent of used file system space at which the sender will stop queueing. When we will reach that percentage, the file system in which the queue is stored will drop all new logs until the percentage of used space drops below that threshold. Set to -1 to never stop processing new logs |
gcPersistedQueueFilesIntervalSeconds | 30 | How often the disk queue should clean sent logs from disk |
bufferDir(deprecated, use queueDir) | System.getProperty("java.io.tmpdir") | Where the appender should store the queue |
queueDir | System.getProperty("java.io.tmpdir") | Where the appender should store the queue |
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
public class LogzioLog4j2Example {
public static void main(String[] args) {
Logger logger = LogManager.getLogger(LogzioLog4j2Example.class);
logger.info("Testing logz.io!");
logger.warn("Winter is coming");
}
}
Each key value you will add to MDC will be added to each log line as long as the thread alive. No further configuration needed.
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import org.apache.logging.log4j.ThreadContext;
public class LogzioLog4j2Example {
public static void main(String[] args) {
Logger logger = LogManager.getLogger(LogzioLog4j2Example.class);
ThreadContext.put("Key", "Value");
logger.info("This log will hold the MDC data as well");
}
}
Will send a log to Logz.io that looks like this:
{
"message": "This log will hold the MDC data as well",
"Key": "Value",
... (all other fields you used to get)
}
Markers are named objects used to enrich log statements, so each log line will be enriched with its own. No further configuration needed.
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import org.apache.logging.log4j.Marker;
import org.apache.logging.log4j.MarkerManager;
public class LogzioLog4j2Example {
public static void main(String[] args) {
Logger logger = LogManager.getLogger(LogzioLog4j2Example.class);
Marker marker = MarkerManager.getMarker("Fatal");
logger.error(marker, "This line has a fatal error");
}
}
Will send a log to Logz.io that looks like this:
{
"message": "This line has a fatal error",
"Marker": "Fatal",
... (all other fields you used to get)
}
- Clone the repository:
git clone https://github.com/logzio/logzio-log4j2-appender.git
cd logzio-log4j2-appender
- Build and run tests:
mvn clean compile
mvn test
- 2.1.0
- Updated LogzioSender version to
2.1.0
- Upgrade packages version
- Upgrade packages version
- Updated LogzioSender version to
- 2.0.1
- Updated LogzioSender version to
2.0.1
- Add
User-Agent
header with logz.io information
- Add
- Updated LogzioSender version to
- 2.0.0 - THIS IS A SNAPSHOT RELEASE - SUPPORTED WITH JDK 11 AND ABOVE
- Updated LogzioSender version to
2.0.0
:- Fixes an issue where DiskQueue was not clearing disk space when using JDK 11 and above.
- Updated LogzioSender version to
- 1.0.19
- Updated LogzioSender version to
1.1.8
:- Fix an issue where log is not being truncated properly between size of 32.7k to 500k.
- Updated LogzioSender version to
- 1.0.18
- updated logzio sender version, fixing IndexOutOfBounds exception with bigqueue
Expand to check old versions
- 1.0.16
- Added exceedMaxSizeAction parameter for handling oversized logs.
- 1.0.15
- Bump versions of
log4j-core
- Bump versions of
- 1.0.14
- Bump versions of
log4j-api
andlog4j-core
- Bump versions of
- 1.0.13
- Fix for issue #38, thanks to @idachev
- Bump versions of
log4j
andguava
- 1.0.11
- add in memory queue option
- change bufferDir(deprecated) to queueDir
- 1.0.10
- Allow to set type using environment variables.
- 1.0.9
- fix guava shaded dependency.
- 1.0.8
- added
compressRequests
parameter to enable gzip compression of the logs before they are sent. - added option to inject system property value into additionalFields, logzioUrl and token.
- added
- 1.0.6 - 1.0.7
- 1.0.5
- 1.0.3 - 1.0.4
- add error message about reason of 400(BAD REQUEST)
- 1.0.2
- Fix problem where logzioToken and logzioUrl could not be environment variables
- 1.0.1
- Fixed an issue: LogzioAppender does not let the JVM exit
- 1.0.0
- Initial releases
- Fork
- Code
mvn test
- Issue a PR :)