There is a number of possible improvements to the architecture of SUSE Manager / Salt integration that should be investigated in order to improve the reliability and scalability of the backend:

  1. Actions are currently scheduled in the minions using the schedule module of Salt. This brings problems with reliability as for instance a minion can be down at the specified schedule time which leads to actions not being executed. Scalability can be an issue as actions being scheduled for many minions might return results to the server at the same time. Instead it might be better to keep control over scheduled actions on the server to allow batching of actions as well as downtimes of minions or even the server. There is a work in progress branch to get started.
  2. For receiving action results we are currently relying on a websocket connection to the Salt event bus in order to receive job return events. This is problematic as the connection might be interrupted leading to the server missing events. Instead we could make use of a master-side returner to write the action results directly into the postgresql database. This would further allow to have setups with multiple Salt masters returning job results to the same database.

Looking for mad skills in:

susemanager salt backend java python postgresql

This project is part of:

Hack Week 15

Activity

  • almost 2 years ago: dvosburg liked SUSE Manager / Salt integration revisited
  • about 2 years ago: whdu liked SUSE Manager / Salt integration revisited
  • about 2 years ago: moio joined SUSE Manager / Salt integration revisited
  • about 2 years ago: moio liked SUSE Manager / Salt integration revisited
  • about 2 years ago: Johannes Renner added keyword "postgresql" to SUSE Manager / Salt integration revisited
  • Show History

    Comments

    Be the first to comment!

    Similar Projects

    MirrorPinky by darix

    *A web frontend for the mirrors in the mirrorbr...


    openSUSE Infrastructure "Factory first"-like policy by tampakrap

    The SLE15 development model follows the Factory...


    Go async (and non-blocking) with HTTP requests by Johannes Renner

    There is a couple of libraries available for as...


    Enhance Staging Project process: reduce the gap between Letter staging and ADI staging by mlin7442

    We have a known defect exists in Staging Projec...


    Telegram to RSS/Atom gate by bbobrov

    Telegram is a proprietary messenger that gained...