LinuxQuestions.org - Logging via stdin

- Linux - Server (https://www.linuxquestions.org/questions/linux-server-73/)

- - Logging via stdin (https://www.linuxquestions.org/questions/linux-server-73/logging-via-stdin-527771/)

Logging via stdin

Is there a logging / log rotation tool which allows me to pipe output into it and it will log it in a particular way? Basically what I'm looging for would have an interface something like the following

Code:

$ ./run_my_tool | stdin2log --path=/path/to/logs --rotate=10 --compress

or maybe

Code:

$ ./run_my_tool | stdin2log --file=/path/to/logs/number%dfile --rotate=10 --compress=bz2

I have looked at logrotate and it doesn't really handle logging output from standard out if the programs you want to log can be run more than once a day. I have also thought about using the append operator (>>) but I want to keep the logs for each individual execution seperate. I have had a good look on Sourceforge and Freshmeat but cannot find anything like it. I can't believe no one has thought of this before, I can only think there is a reason why it is a bad idea. If no one knows of a tool like this and people think it's a good idea I'll happily implement it myself, I just don't want to go re-inventing the wheel.

Code:

#!/bin/bash

# Purpose: Catch stdin and archive when done.

# Deps: Bash, GNU utils

# set -xe

umask 027; progn=${0//*\//}; progn=${progn%.sh}

log_name="${progn}"; log_base="/var/log"

log_stamp=`/bin/date '+%Y%m%d_%H%M%S%N' 2>/dev/null`

log_file="${log_base}/${log_name}.${log_stamp}"



help() { # Echo some help lines here.

 echo "${progn}: -c command and arguments -l logfile"; exit 1; }



loglines() { # Formatting and logging

 while read line; do line="$(/bin/date "+%b %d %H:%M:%S:%N" 2>/dev/null) \

 $(hostname -s 2>/dev/null) ${progn} [$$]: ${line:="EMPTY LINE"}"

 echo "${line}" 2>&1>> "${log_file}"; done

 [ $PIPESTATUS -eq 1 ] && gzip "${log_file}" 2>/dev/null; }



[ $# -eq 0 ] && help; [ $# -gt 64 ] && help

[ "${1:0:1}" != "-" -o "${1:0:2}" = "--" ] && help

while getopts c:l:h OPT; do case "${OPT}" in

  c) [ -n "${OPTARG}" ] && cmdline="${OPTARG}";;

  l) log_file="${OPTARG}";;

  *|h) help;; esac

done; exec $cmdline 2>&1 | loglines

exit 0

Execute as

Code:

stdin2log -c "seq 1 100" -l /var/log/name.log

YMMV(VM), as usual ;-p

Very cool script! It's not quite what I'm looking for but it certainly shown me that implementing what I need will be a hell of a lot simpler than I'd imagined. Thanks.

Post your script when you're done?

This is what I've put together, much easier that what I had first envisioned. Thanks for your help unSpawn.

Code:

#!/bin/bash



umask 027;

rotate=10; path='';



# Fail and print the usage

help() {

    if [ $# -gt 0 ]

    then

        echo "error [$1]";

    fi

    echo "stdin2log -r [number of times to rotate the log]"\

      " -p [path to log directory or filename template]";

    exit 1;

}



# Build the filename

getFilename() {

    if [ $1 = 'file' ]

    then

        echo -n $(printf $2 $3);

    else

        echo -n "${2}log${3}";

    fi;

}



while getopts r:p:h: OPT;

do

    case "${OPT}" in

        r) [ ${OPTARG} -gt 0 ] && rotate=${OPTARG};;

        p) path="${OPTARG}";;

        *|h) help;;

    esac

done;



# Figure out whether it's a directory or a filename template

[ -z $path ] && help "Empty path";

if $(echo $path | grep -q '\/$')

then

    pathtype='dir';

elif $(echo $path | grep -q '%d' )

then

    pathtype='file';

else

    help "Invalid path ${path}";

fi;

# Run through the logs backwards moving them back

for i in $(seq $(($rotate-1)) 1)

do

    log_file=$(getFilename $pathtype $path $i)

    if [ -e $log_file ]

    then

        mv -f $log_file $(getFilename $pathtype $path $(($i+1)))

    fi

done;



# Write all lines from stdin to the first file

log_file=$(getFilename $pathtype $path 1);

touch $log_file;

while read line;

do

    echo "${line}" >> "${log_file}" 2>&1

done;