hhm0 / nbudstee Goto Github PK

Tees Input to zero or more non-blocking Unix domain sockets, each of which can have zero or more connected readers.
Also copies to STDOUT, unless -n/--no-stdout is used.
Input defaults to STDIN

nbudstee [options] uds1 uds2 ...

Where uds1 uds2 ... are zero or more filenames to use as Unix domain sockets.

• -n, --no-stdout
  Do not copy input to STDOUT.
• -b, --unlink-before
  First try to unlink any existing sockets. This will not try to unlink non-sockets.
• -u, --unlink-after
  Try to unlink all sockets when done.
• -m, --max-queue bytes
  Maximum amount of data to buffer for each connected reader (approximate).
  Accepts suffixes: k, M, G, for kilobytes, megabytes, gigabytes (multiples of 1024).
  Default: 64k.
  Above this limit new data for that socket reader will be discarded.
• -i, --input file
  Read from file instead of STDIN.
• -I, --input-reopen file
  Read from file instead of STDIN.
  When the end of input is reached, reopen from the beginning.
  This is primarily intended for FIFOs.

On demand logging, in particular in shell pipelines.

... | process | nbudstee -u uds1 | process | ...
... | process 2> >(nbudstee -n -u uds2) | ...

Any number of readers can read from each socket, on demand, without blocking the main pipeline, or requiring data to be dumped to a file.
This is useful for occasional debugging/sampling, or for inserting/removing tees without interrupting the main pipeline.

A possible command to read from a Unix domain socket is:

nc -U uds1

• No attempt is made to line-buffer or coalesce the input. A reader could receive a partial line when connecting/disconnecting if the input source is not line-buffered.
  If required line-buffer the input before piping it to nbudstee.
• Writes to STDOUT are blocking.

• This project is hosted at https://github.com/JGRennison/nbudstee
• A Ubuntu PPA is currently available at https://launchpad.net/~j-g-rennison/+archive/nbudstee

GPLv2