[Openstack] [Neutron][LBaaS] Random 503 returned by Haproxy

Zuo Changqian dummyhacker85 at gmail.com
Thu May 8 06:52:43 UTC 2014


Recently I am working on LBaaS service. When I set up a pool and it serves,
the haproxy process randomly returns 503:

503 Service Unavailabe

No server is available to handle this request

This could be easily reproduced and I am pretty sure when this happened,
member servers are up.

Meanwhile, I ran another haproxy process, started with CentOS init script.
And the problem never showed. I replaced /etc/haproxy/haproxy.cfg with
.../lbaas/{poo_id}/conf file (change stat socket path and group) and it
still OK.

So, I checked init script and LBaaS-agent code, found there is a "-D"
argument in the init script when creating Haproxy process. I killed haproxy
process for the pool, and restart it with one more "-D" argument in pool's
namespace, and the problem finally gone.

I don't know much about Haproxy, how could this "-D" argument make such
difference (since there "deamon" options in the configuration file) ?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140508/5829a3d3/attachment.html>

More information about the Openstack mailing list