pmie_daily is intended to be run once per day, preferably in the early morning, as soon after midnight as practicable. Its task is to rotate the log files for the running pmie processes - these files may grow without bound if the ``print'' action is used, or any other pmie action writes to its stdout/stderr streams. After some period, old pmie log files are discarded. This period is 14 days by default, but may be changed using the -k option. Two special values are recognized for the period ( discard ), namely 0 to keep no log files beyond the current one, and forever to prevent any log files being discarded.
Log files can optionally be compressed after some period ( compress ), to conserve disk space. This is particularly useful for large numbers of pmie processes under the control of pmie_check . The -x option specifies the number of days after which to compress archive data files, and the -X option specifies the program to use for compression - by default this is xz (1). Use of the -Y option allows a regular expression to be specified causing files in the set of files matched for compression to be omitted - this allows only the data file to be compressed, and also prevents the program from attempting to compress it more than once. The default regex is ".(meta|index|Z|gz|bz2|zip|xz|lzma|lzo|lz4)$" - such files are filtered using the -v option to egrep (1).
Use of the -m option causes pmie_daily to construct a summary of the log files generated for all monitored hosts in the last 24 hours (lines matching `` OK '' are culled), and e-mail that summary to the set of space-separated addresses .
pmie_check may be run at any time, and is intended to check that the desired set of pmie (1) processes are running, and if not to re-launch any failed inference engines. Use of the -s option provides the reverse functionality, allowing the set of pmie processes to be cleanly shutdown. Use of the -C option queries the system service runlevel information for pmie , and uses that to determine whether to start processes.
Both pmie_check and pmie_daily are controlled by PCP inference engine control file(s) that specify the pmie instances to be managed. The default control file is $PCP_PMIECONTROL_PATH but an alternate may be specified using the -c option. If the directory $PCP_PMLOGGERCONTROL_PATH .d (or control .d from the -c option) exists, then the contents of any additional control files therein will be appended to the main control file (which must exist).
Warning : The $PCP_PMIECONTROL_PATH and $PCP_PMIECONTROL_PATH .d files must not be writable by any user other than root.
The control file(s) should be customized according to the following rules that define for the current version (1.1) of the control file format.
host y|n y|n logfile args
The following sample control lines specify one pmie instance monitoring the local host ( wobbly ), and another monitoring performance metrics from the host splat .
wobbly n PCP_LOG_DIR/pmie/wobbly -c config.default splat n PCP_LOG_DIR/pmie/splat -c splat/cpu.conf
Typical crontab (5) entries for periodic execution of pmie_daily and pmie_check are given in $PCP_SYSCONF_DIR/pmie/crontab (unless installed by default in /etc/cron.d already) and shown below.
# daily processing of pmie logs 08 0 * * * $PCP_BINADM_DIR/pmie_daily # every 30 minutes, check pmie instances are running 28,58 * * * * $PCP_BINADM_DIR/pmie_check
In order to ensure that mail is not unintentionally sent when these scripts are run from cron (8) diagnostics are always sent to log files. By default, these files are $PCP_LOG_DIR/pmie/pmie_daily.log and $PCP_LOG_DIR/pmie/pmie_check.log but this can be changed using the -l option. If this log file already exists when the script starts, it will be renamed with a .prev suffix (overwriting any log file saved earlier) before diagnostics are generated to the new log file.
The output from the cron execution of the scripts may be extended using the -V option to the scripts which will enable verbose tracing of their activity. By default the scripts generate no output unless some error or warning condition is encountered.
The -N option enables a ``show me'' mode, where the actions are echoed, but not executed, in the style of ``make -n''. Using -N in conjunction with -V maximizes the diagnostic capabilities for debugging.