plainbox-session-structure (7)

Synopsis

This page documents the structure of the Plainbox per-session directory.

Description

Each session is represented by a directory. Typically all sessions are stored in the $XDG_CACHE_HOME/plainbox/sessions/ directory. Each directory there is a randomly-named session comprised of the following files and directories.

session:

A state with the serialized state of the session. Currently it is a JSON document compressed with the gzip compression scheme. You can preview the contents of this file with zcat session | json_pp where zcat (1) and json_pp` (1) are external system utilities.

The session file stores the state of the session. State is represented by several structures which are further documented in plainbox-session-state. This file is essential for resuming a session but is also useful for debugging.

io-logs:

A directory with files representing input-output operations performed by particular jobs. There are three files for each job. One for Plainbox itself and two more for human-readable debugging. The files are:

*.record.gz:

A file internal to Plainbox, containing representation of all of the input-output operations performed by the specified job definition’s command process.

The format for this file is a gzip-compressed sequence of records, represented as separate lines, terminated with the newline character. Each record is a small JSON list of exactly three elements. The first element is a JSON number representing the delay since the previous element was generated OR the delay before the process startup time, for the first record. The second name is the name of the communication stream. Currently only stdout and stderr are used. The third and last element of each record is a base-64 encoded binary string representing the communication that took place.

The leading part of the filename is currently the identifier of the job definition but this is subject to change to allow for multiple log files associated with a single job in a given session.

To figure out which log file is associated with each job definition, refer to the state file (session).

*.stdout:
Plain-text representation of the entire stdout stream as it was printed by the command process. This file is purely for debugging and is ignored by Plainbox. It may cease to be generated at some future time.
*.stderr:
Similarly to .stdout but for the stderr stream.
CHECKBOX_DATA:
A directory associated with the PLAINBOX_SESSION_SHARE (7) per-session runtime directory where jobs may deposit files to perform a primitive form of IJC (inter-job-communication).
comments powered by Disqus