Options for logging resource usage metrics.

If a crew_options_metrics() object is supplied to the options_metrics argument of a crew controller function, then the autometric R package will record resource usage metrics (such as CPU and memory usage) for each running worker. Logging happens in the background (through a detached POSIX) so as not to disrupt the R session. On Unix-like systems, crew_options_metrics() can specify /dev/stdout or /dev/stderr as the log files, which will redirect output to existing logs you are already using. autometric::log_read() and autometric::log_plot() can read and visualize resource usage data from multiple log files, even if those files are mixed with other messages.

Usage

crew_options_metrics(path = NULL, seconds_interval = 5)

Arguments

path

Where to write resource metric log entries for workers. path = NULL disables logging. path equal to "/dev/stdout" (or "/dev/stderr") sends log messages to the standard output (or standard error) streams, which is recommended on Unix-like systems because then output will go to the existing log files already configured for the controller, e.g. through crew_options_local() in the case of crew_controller_local(). If path is not NULL, "/dev/stdout", or "/dev/stderr", it should be a directory path, in which case each worker instance will write to a new file in that directory.

After running enough tasks in crew, you can call autometric::log_read(path) to read all the data from all the log files in the files or directories at path, even if the logs files are mixed with other kinds of messages. Pass that data into autometric::log_read() to visualize it.

seconds_interval

Positive number, seconds between resource metric log entries written to path.

Value

A classed list of options for logging resource usage metrics.

See also

Other options: crew_options_local()

Examples

  crew_options_metrics()
#> $path
#> NULL
#> 
#> $seconds_interval
#> [1] 5
#> 
#> attr(,"class")
#> [1] "crew_options_metrics" "crew_options"