What is it?

A kind of tee that writes stdin to a rotated log file(s).

I got tired of the absence of lightweight log rotation utility. logrotate requires a whole infrastructure around itself. bash-only piping requires more than a one-liner, plus bash has to do the read-write loop, which doesn't feel optimal. Or, worse, bash script needs to restart the main program.

Too late I discovered multilog in this great SuperUser answer.

Examples

verbosecommand | logrotee /var/log/verbosecommand.log \
> /dev/null

or

verbosecommand | logrotee \
  --compress "bzip2 {}" --compress-suffix .bz2 \
  --null --chunk 2M \
  /var/log/verbosecommand.log

If the output chunk file + <suffix> already exists, add an unique sequential suffix, currently .1, .2, .3, you get the idea.

Usage

logrotee [-z|--compress <compress-command>] [-s|--compress-suffix <suffix>] [-n|--null] [--chunk|-k <size>] <name>

• <name>: Log file name. If the file exists, its name will be appended a number, like: /var/log/verbosecommand.log.1, /var/log/verbosecommand.log.2, and so on. If --compress-suffix "bz2" is given, will check for /var/log/verbosecommand.log.bz2, /var/log/verbosecommand.log.1.bz2 and so on.
• -z, --compress <compress-command> Optional parameter - command to run on a complete chunk file, typically packer. {} is substituted with a full file path. It's the packer's responsibility to delete the original chunk file.
• -s, --compress-suffix <suffix> File suffix to append to check for <name> existence, if any.
• -n, --null Don't print the stdin to stdout. Not sure if it gains anything compared to >/dev/null.
• -k, --chunk <chunk-size> Chunk file size in bytes, after which it's closed and maybe compressed. Default is ~10Mb.

Planned:

• -d, --dates Use dates instead of sequential numbers for rotation suffix.
• Rotate by time instead of size.
• -m, --max-files <amount> Maximum amount of rotated files. Delete the oldest chunk files when over this limit.
• Does not check if the files were not created by this instance of logrotee.
• -z shortcut for gzip/bz2.
• Handle signals.
• Parse k/M/etc suffixes in file sizes.
• Handle out-of-space situation.
• Do not launch a compressor until the previous exits.
• Looks like some people need to prepend log lines with timestamps.