The syslog-ng PE application sends the log messages to the official HDFS client library, which forwards the data to the HDFS nodes. The way how syslog-ng PE interacts with HDFS is described in the following steps.
After syslog-ng PE is started and the first message arrives at the hdfs destination, the hdfs destination tries to connect to the HDFS NameNode. If the connection fails, syslog-ng PE will repeatedly attempt to connect again after the period set in time-reopen() expires.
syslog-ng PE checks if the path to the logfile exists. If a directory does not exist syslog-ng PE automatically creates it. syslog-ng PE creates the destination file (using the filename set in the syslog-ng PE configuration file, with a UUID suffix to make it unique, for example, /usr/hadoop/logfile.txt.3dc1c59e-ab3b-4b71-9e81-93db477ed9d9) and writes the message into the file. After the file is created, syslog-ng PE will write all incoming messages into the hdfs destination.
NOTE: When the hdfs-append-enabled() option is set to true, syslog-ng PE will not assign a new UUID suffix to an existing file, because it is then possible to open a closed file and append data to that.
NOTE: You cannot set when log messages are flushed. Hadoop performs this action automatically, depending on its configured block size, and the amount of data received. There is no way for the syslog-ng PE application to influence when the messages are actually written to disk. This means that syslog-ng PE cannot guarantee that a message sent to HDFS is actually written to disk. When using flow-control, syslog-ng PE acknowledges a message as written to disk when it passes the message to the HDFS client. This method is as reliable as your HDFS environment.
If the HDFS client returns an error, syslog-ng PE attempts to close the file, then opens a new file and repeats sending the message (trying to connect to HDFS and send the message), as set in the retries() parameter. If sending the message fails for retries() times, syslog-ng PE drops the message.
The syslog-ng PE application closes the destination file in the following cases:
syslog-ng PE is reloaded
syslog-ng PE is restarted
The HDFS client returns an error.
If the file is closed and you have set an archive directory, syslog-ng PE moves the file to this directory. If syslog-ng PE cannot move the file for some reason (for example, syslog-ng PE cannot connect to the HDFS NameNode), the file remains at its original location, syslog-ng PE will not try to move it again.
The syslog-ng PE application is also compatible with MapR File System (MapR-FS), starting from version 5.4, syslog-ng Premium Edition is MapR certified. MapR-FS provides better performance, reliability, efficiency, maintainability, and ease of use compared to the default Hadoop Distributed Files System (HDFS). To use MapR-FS with syslog-ng PE, complete the following steps:
Install MapR libraries. Instead of the official Apache HDFS libraries, MapR uses different libraries. The supported version is MapR 4.x.
Download the libraries from the Maven Repository and Artifacts for MapR or get it from an already existing MapR installation.
Install MapR. If you do not know how to install MapR, follow the instructions on the MapR website.
In a default MapR installation, the required libraries are installed in the following path: /opt/mapr/lib.
Enter the path where MapR was installed in the class-path option of the hdfs destination, for example:
If the libraries were downloaded from the Maven Repository, the following additional libraries will be requiered. Note that the version numbers in the filenames can be different in the various Hadoop releases:commons-collections-3.2.1.jar, commons-logging-1.1.3.jar, hadoop-auth-2.5.1.jar, log4j-1.2.15.jar, slf4j-api-1.7.5.jar, commons-configuration-1.6.jar, guava-13.0.1.jar, hadoop-common-2.5.1.jar, maprfs-4.0.2-mapr.jar, slf4j-log4j12-1.7.5.jar, commons-lang-2.5.jar, hadoop-0.20.2-dev-core.jar, json-20080701.jar, protobuf-java-2.5.0.jar, zookeeper-3.4.5-mapr-1406.jar.
Configure the hdfs destination in syslog-ng PE.
Example: Storing logfiles with MapR-FS
The following example defines an hdfs destination for MapR-FS using only the required parameters.
Version 7.0.3 and later supports Kerberos authentication to authenticate the connection to your Hadoop cluster. syslog-ng PE assumes that you already have a Hadoop and Kerberos infrastructure.
NOTE: If you configure Kerberos authentication for a hdfs() destination, it affects all hdfs() destinations. Kerberos and non-Kerberos hdfs() destinations cannot be mixed in a syslog-ng PE configuration. This means that if one hdfs() destination uses Kerberos authentication, you have to configure all other hdfs() destinations to use Kerberos authentication too.
Failing to do so results in non-Kerberos hdfs() destinations being unable to authenticate to the HDFS server.
NOTE: If you want to configure your hdfs() destination to stop using Kerberos authentication, namely, to remove Kerberos-related options from the hdfs() destination configuration, make sure to restart syslog-ng PE for the changes to take effect.
You have configured your Hadoop infrastructure to use Kerberos authentication.
You have a keytab file and a principal for the host running syslog-ng PE.
You have installed and configured the Kerberos client packages on the host running syslog-ng PE. (That is, Kerberos authentication works for the host, for example, from the command line using the kinit user@REALM -k -t <keytab_file> command.)
The hdfs destination stores the log messages in files on the Hadoop Distributed File System (HDFS). The hdfs destination has the following options.
The following options are required: hdfs-file(), hdfs-uri(). Note that to use hdfs, you must add the following lines to the beginning of your syslog-ng PE configuration:
The syslog-ng PE module directory: /opt/syslog-ng/lib/syslog-ng/java-modules/
Description: The list of the paths where the required Java classes are located. For example, class-path("/opt/syslog-ng/lib/syslog-ng/java-modules/:/opt/my-java-libraries/libs/"). If you set this option multiple times in your syslog-ng PE configuration (for example, because you have multiple Java-based destinations), syslog-ng PE will merge every available paths to a single list.
For the hdfs destination, include the path to the directory where you copied the required libraries (see Prerequisites), for example, client-lib-dir("/opt/syslog-ng/lib/syslog-ng/java-modules/:/opt/hadoop/libs/").
Description: This option enables putting outgoing messages into the disk-buffer file of the destination to avoid message loss in case of a system failure on the destination side. It has the following options:
Description: Defines the folder where the disk-buffer files are stored.
Note that changing the value the dir() option will not move or copy existing files from the old directory to the new one.
When creating a new dir() option for a disk-buffer file, or modifying an existing one, make sure you delete the persist file.
syslog-ng PE creates disk-buffer files based on the path recorded in the persist file. Therefore, if the persist file is not deleted after modifying the dir() option, then following a restart, syslog-ng PE will look for or create disk-buffer files in their old location. To ensure that syslog-ng PE uses the new dir() setting, the persist file must not contain any information about the destinations which the disk-buffer file in question belongs to.
Description: This is a required option. The maximum size of the disk-buffer file in bytes. The minimum value is 1048576 bytes. If you set a smaller value, the minimum value will be used automatically. It replaces the old log-disk-fifo-size() option.
Description: Use this option if the option reliable() is set to no. This option contains the number of messages stored in overflow queue. It replaces the old log-fifo-size() option. It inherits the value of the global log-fifo-size() option if provided. If it is not provided, the default value is 10000 messages. Note that this option will be ignored if the option reliable() is set to yes.
Description: Use this option if the option reliable() is set to yes. This option contains the size of the messages in bytes that is used in the memory part of the disk-buffer file. It replaces the old log-fifo-size() option. It does not inherit the value of the global log-fifo-size() option, even if it is provided. Note that this option will be ignored if the option reliable() is set to no.
Description: The number of messages stored in the output buffer of the destination. Note that if you change the value of this option and the disk-buffer file already exists, the change will take effect when the disk-buffer file becomes empty.
Description: If set to yes, syslog-ng PE cannot lose logs in case of reload/restart, unreachable destination or syslog-ng PE crash. This solution provides a slower, but reliable disk-buffer option. It is created and initialized at startup and gradually grows as new messages arrive. If set to no, the normal disk-buffer option will be used. This provides a faster, but less reliable disk-buffer option.
Caution: Hazard of data loss!
If you change the value of reliable() option when there are messages in the disk-buffer file, the messages stored in the disk-buffer file will be lost.
number (for percentage) between 0 and 1
Description: Limits the truncation of the disk-buffer file. Truncating the disk-buffer file can slow down disk I/O operations, but it saves disk space. As a result, syslog-ng PE only truncates the file if the possible disk gain is more than truncate-size-ratio() times disk-buf-size().
One Identity recommends that you do not modify the value of the truncate-size-ratio() option unless you are fully aware of the potential performance implications.
Example: Examples for using disk-buffer()
In the following case, reliable disk-buffer() is used.
Description: The syslog-ng PE application can store fractions of a second in the timestamps according to the ISO8601 format. The frac-digits() parameter specifies the number of digits stored. The digits storing the fractions are padded by zeros if the original timestamp of the message specifies only seconds. Fractions can always be stored for the time the message was received.
NOTE: The syslog-ng PE application can add the fractions to non-ISO8601 timestamps as well.
NOTE: As syslog-ng PE is precise up to the microsecond, when the frac-digits() option is set to a value higher than 6, syslog-ng PE will truncate the fraction seconds in the timestamps after 6 digits.
true | false
Description: When hdfs-append-enabled is set to true, syslog-ng PE will append new data to the end of an already existing HDFS file. Note that in this case, archiving is automatically disabled, and syslog-ng PE will ignore the hdfs-archive-dir option.
When hdfs-append-enabled is set to false, the syslog-ng PE application always creates a new file if the previous has been closed. In that case, appending data to existing files is not supported.
When you choose to write data into an existing file, syslog-ng PE does not extend the filename with a UUID suffix because there is no need to open a new file (a new unique ID would mean opening a new file and writing data into that).
Before enabling the hdfs-append-enabled option, ensure that your HDFS server supports the append operation and that it is enabled. Otherwise syslog-ng PE will not be able to append data into an existing file, resulting in an error log.
Description: The path where syslog-ng PE will move the closed log files. If syslog-ng PE cannot move the file for some reason (for example, syslog-ng PE cannot connect to the HDFS NameNode), the file remains at its original location. For example, hdfs-archive-dir("/usr/hdfs/archive/").
NOTE: When hdfs-append-enabled is set to true, archiving is automatically disabled, and syslog-ng PE will ignore the hdfs-archive-dir option.
Description: The path and name of the log file. For example, hdfs-file("/usr/hdfs/mylogfile.txt"). syslog-ng PE checks if the path to the logfile exists. If a directory does not exist syslog-ng PE automatically creates it.
hdfs-file() supports the usage of macros. This means that syslog-ng PE can create files on HDFS dynamically, using macros in the file (or directory) name.
NOTE: When a filename resolved from the macros contains a character that HDFS does not support, syslog-ng PE will not be able to create the file. Make sure that you use macros that do not contain unsupported characters.
Example: Using macros in filenames
In the following example, a /var/testdb_working_dir/$DAY-$HOUR.txt file will be created (with a UUID suffix):
As an example, if it is the 31st day of the month and it is 12 o'clock, then the name of the file will be 31-12.txt.
Description: The maximum length of the filename. This filename (including the UUID that syslog-ng PE appends to it) cannot be longer than what the file system permits. If the filename is longer than the value of hdfs-max-filename-length, syslog-ng PE will automatically truncate the filename. For example, hdfs-max-filename-length("255").
Description: The list of Hadoop resources to load, separated by semicolons. For example, hdfs-resources("/home/user/hadoop/core-site.xml;/home/user/hadoop/hdfs-site.xml").
Description: The URI of the HDFS NameNode is in hdfs://IPaddress:port or hdfs://hostname:port format. When using MapR-FS, the URI of the MapR-FS NameNode is in maprfs://IPaddress or maprfs://hostname format, for example: maprfs://10.140.32.80. The IP address of the node can be IPv4 or IPv6. For example, hdfs-uri("hdfs://10.140.32.80:8020"). The IPv6 address must be enclosed in square brackets () as specified by RFC 2732, for example, hdfs-uri("hdfs://[FEDC:BA98:7654:3210:FEDC:BA98:7654:3210]:8020").
Description: Specify the Java Virtual Machine (JVM) settings of your Java destination from the syslog-ng PE configuration file.
You can set this option only as a global option, by adding it to the options statement of the syslog-ng configuration file.
Description: The path to the Kerberos keytab file that you received from your Kerberos administrator. For example, kerberos-keytab-file("/opt/syslog-ng/etc/hdfs.headless.keytab"). This option is needed only if you want to authenticate using Kerberos in Hadoop. You also have to set the hdfs-option-kerberos-principal() option. For details on the using Kerberos authentication with the hdfs() destination, see Kerberos authentication with syslog-ng hdfs() destination.
Use the global setting (which defaults to drop-message)
Description: Controls what happens when type-casting fails and syslog-ng PE cannot convert some data to the specified type. By default, syslog-ng PE drops the entire message and logs the error. Currently the value-pairs() option uses the settings of on-error().
drop-message: Drop the entire message and log an error message to the internal() source. This is the default behavior of syslog-ng PE.
drop-property: Omit the affected property (macro, template, or message-field) from the log message and log an error message to the internal() source.
fallback-to-string: Convert the property to string and log an error message to the internal() source.
silently-drop-message: Drop the entire message silently, without logging the error.
silently-drop-property: Omit the affected property (macro, template, or message-field) silently, without logging the error.
silently-fallback-to-string: Convert the property to string silently, without logging the error.
number [of attempts]
Description: The number of times syslog-ng PE attempts to send a message to this destination. If syslog-ng PE could not send a message, it will try again until the number of attempts reaches retries(), then drops the message.
A format conforming to the default logfile format.
Description: Specifies a template defining the logformat to be used in the destination. Macros are described in Macros of syslog-ng PE. Please note that for network destinations it might not be appropriate to change the template as it changes the on-wire format of the syslog protocol which might not be tolerated by stock syslog receivers (like syslogd or syslog-ng itself). For network destinations make sure the receiver can cope with the custom format defined.
Description: Sets the maximum number of messages sent to the destination per second. Use this output-rate-limiting functionality only when using the disk-buffer option as well to avoid the risk of losing messages. Specifying 0 or a lower value sets the output limit to unlimited.
Description: The time to wait in seconds before an idle destination file is closed. Note that if hdfs-archive-dir option is set and time-reap expires, archiving is triggered for the affected file.
name of the timezone, or the timezone offset
Description: Convert timestamps to the timezone specified by this option. If this option is not set, then the original timezone information in the message is used. Converting the timezone changes the values of all date-related macros derived from the timestamp, for example, HOUR. For the complete list of such macros, see Date-related macros.
The timezone can be specified as using the name of the (for example, time-zone("Europe/Budapest")), or as the timezone offset in +/-HH:MM format (for example, +01:00). On Linux and UNIX platforms, the valid timezone names are listed under the /usr/share/zoneinfo directory.
rfc3164, bsd, rfc3339, iso
Description: Override the global timestamp format (set in the global ts-format() parameter) for the specific destination. For details, see ts-format().