The role of this application is to provide the ability to easily benchmark the file input capabilities of various log shipper clients to a Kafka output destination. Metrics from the log shippers are collected via metricbeat.
- Must be running Go >= 1.8.x, although was built with 1.10.0
- Ensure
GOPATHis set to$HOME/goand that the directory exists - Ensure
$HOME/go/binis part of youPATHenvironment variable - Require go vendoring library:
github.com/kardianos/govendor(go get github.com/kardianos/govendor)
- Fluent Bit: v0.13.0
- Filebeat : v6.1.1
- Logstash : v6.1.1
- Rsyslogd : v8.34.0
- github.com/kardianos/govendor
- github.com/c9s/goprocinfo/linux
- github.com/Pallinder/go-randomdata
- github.com/mitchellh/go-ps
- Metricbeat: v6.1.1
make deps
make clean
make all
The config, which is in JSON format, should contain the following fields:
additional_metricbeat_fields: An object consisting of additional key/value properties to add the the metricbeat data. (Type: map[string]string, Default: )custom_log_entry: If set, the this specific log entry will be written to the files instead of a randomly generated one. (Type: string, Default: )enable_random: If set to true, the application will randomly choose a line size and wait time between writes. (Type: boolean, Default: false)kafka_broker_list: List of Kafka broker hostnames (HOST:PORT) to use in the log shippers. Currently only this output destination is supported. (Type: []string, Default)log_files_base_dir: Location where the sample log files will be created (Type: string, Default: )log_line_size: The size (character length) of the log entry to be randomly generated. (Type: int, Default: 50)log_shipper_bin_path: The path to the log shipper binary. (Type: string, Default: )log_shipper_flags: The flags to use when executing the log shipper binary. (Type: string, Default: )log_shipper_name: The name of the log shipper. (Type: string, Default: )log_shipper_process_name: The running process name of the log shipper. (Type: string, Default: )max_procs: The max number of processors this benchmarking app should use. (Type: int, Default: )metrics_dir: The directory in which the collected process metrics will be stored. (Type: string, Default: )module_dir: The directory in which the.soshipper module is found. (Type: string, Default: )module_name: The filename of the module excluding the.soextension. (Type: string, Default: )num_active_log_files: The number of active log files that will be written to concurrently/in-parallel. (Type: int, Default: 10)random_line_size: The MIN,MAX range for the length of the line in characters. (Type []int, Default: )random_write_wait: The MIN,MAX range for period (in milliseconds) bewteen writes to each individual log files. (Type []int, Default: )total_run_time_seconds: The total time (in seconds) to run the benchmark. (Type int, Default: )working_dir: The working directory in which the module will be running. (Type: string, Default: )write_wait_period_ms: The period (in milliseconds) bewteen writes to the each individual log files. (Type int, Default: )
Samples can be found in the _sample_configs directory.
The individual shippers work with a plugin based system. In order to benchmark a new shipper, you must create a plugin that respects the following interface
type Shipper interface {
Name()
Run(binPath string, cmdArgs []string, workingDir string, filesToMonitor []string, kafkaBrokers []string, filebeatExec chan *exec.Cmd, terminate chan bool, wg *sync.WaitGroup)
CleanupFiles()
BuildConfig(confDestPath string, filesToMonitor []string, kafkTopicName string, kafkaBrokersList []string)
GetVersion() string
}
To compile the plugin manually, run the following command:
go build -a -v -buildmode=plugin -o output/path/to/plugin.so source/path/to/plugin.go
The benchmarks for a given log shipper are executed as follows:
./logshipper-benchmark [PATH_TO_CONFIG]
To terminate the benchmark, simply hit Ctrl+C. If the process is backgrounded, to initial a clean shutdown, you must kill the logshipper-benchmark with a SIGINT or SIGTERM signal.
For example:
kill -INT [PID]
This application was developed quickly to have the ability to efficiently create benchmarks for a variety of log shippers. Having said that, it's certain this code isn't optimal and may likely contain bugs. If you find any bugs/performance improvements to this application, feel free to create a bug report or issue a PR. If you have any additional log shippers you'd like to test, or even different version of them, a PR for thew new log shipper would be much appreciated.
Alain Lefebvre [email protected]