OpenBSD Journal

High Availability OpenBSD Clusters

Contributed by jose on from the more-works-in-progress dept.

ubiquitin writes: "Srebrenko Sehic in Denmark has been working on HAOS which is a project to build high availability features into OpenBSD clusters. Note that the intent is not load-sharing clustering but concurrent or active-active clusters." Security, redundancy, and availability. This would be a nice combination to have in the back office.

(Comments are closed)

  1. By Anonymous Coward () on

    So these clusters aren't load balancing, they are active-active clusters. What exactly does that mean, and more importantly, what are they used for?

    1. By anders () on

      I'm no authority and might just be stating the obvious, but rather than having all the boxes in a cluster working on a task at once, HAOS would allow boxes to communicate with each other and pass the responsibility of a task along in case of system failure. Sort of a self-healing network.

    2. By Anonymous Coward () on

      Like MS clustering...

      1. By Anonymous Coward () on

        And like all the other clusters used in the corporations to provide HA services

    3. By click46 () on

      its simply fail-over redundancy. no fancy load balancing. if one fails, HAOS redirects everything to another box.

      that is all.

    4. By Srebrenko Sehic () on

      No. HAOS is an active-standby cluter software.
      In general, this means that one node in the cluter is inactive. Sort of a cold standby node.

      However, you can get really fancy and have nodes in a cluster perform different tasks (a web and a database server), but let nodes do a failover to each other. This is what's called mutual cascading/takeover. This way, all your boxes are active, but in case of a failure, you get a performance penality.

  2. By Eduardo Alvarenga () eduardo at thrx dot dyndns dot org on mailto:eduardo at thrx dot dyndns dot org

    The idea is welcome, it'll be nice if it's avaliable. But where is the code? At least the development case is done?

    1. By Srebrenko Sehic () on

      It's still in very eraly stages, but there is code already done. Unfortunetly, I've been quite busy last few months, so nothing new happened.

      People wanted to look at code, but I don't feel anybody whould gain much from what is written so far. Stay tuned.

  3. By Dan () on

    How does it stands with the patent issue?

    1. By Srebrenko Sehic () on

      You are thinking about VRRP patent issues? HAOS will not use a virtual MAC address between clusters nodes, but will advertise new MAC addresses using gratitious ARP. AFAIK, this is not patented. I've seen this in many other places as well.

      Linux-HA uses the same trick to do network failovers. Send an ARP request to the broadcast address and let the nodes on the segment learn the new address. Most OS's will then update the machines ARP cache, and not wait for the ARP cache to timeout.

    2. By Anonymous Coward () on

      In this e-mail Robert Barr from Cisco answers to the main developer of keepalived that Cisco will not attack VRRP implementations unless a patent claim is asserted against Cisco. Unfortunately he also states that he expects that IBM's stance on this is the opposite of Cisco's.

  4. By Anonymous () on

    In the meantime, just for availability, there is a VRRP port to OpenBSD available at:
    I've used it with much success.


Copyright © - Daniel Hartmeier. All rights reserved. Articles and comments are copyright their respective authors, submission implies license to publish on this web site. Contents of the archive prior to as well as images and HTML templates were copied from the fabulous original with Jose's and Jim's kind permission. This journal runs as CGI with httpd(8) on OpenBSD, the source code is BSD licensed. undeadly \Un*dead"ly\, a. Not subject to death; immortal. [Obs.]