[Pulp-list] pulp clustering / HA

[Barnaby Court]

Hi,

With Pulp 2.3.x the scenario you described sounds about right.  The database could be clustered but the Pulp 2.3.x server does not support clustering.  With Pulp 2.4.x (currently in Beta, getting close to Release Candidate) are much closer to supporting true active/active HA.  The primary functionality of the pulp server has been split out into components (web server, generic worker, resource manager, and scheduler) that are hanging off of a message bus.  You can run as many web servers & worker as you like.  Only one resource manager and scheduler can be run at a given time.  A non pulp service could be used to monitor and start new copies of those services if your primary server dies.  More details on the upcoming 2.4 installation & configuration can be found at https://pulp-user-guide.readthedocs.org/en/latest/installation.html.

-Barnaby

----- Original Message -----
From: "Matej Susta" <Matej.Susta at infor.com>
To: pulp-list at redhat.com
Sent: Monday, June 23, 2014 9:35:41 AM
Subject: [Pulp-list] pulp clustering / HA

Hi!

I'm currently trying out pulp with a plan to deploy it in production.
However even though I read all the docs I haven't found any good
information about HA possibilities and backups.

Only thread on this list seems to be:
https://www.redhat.com/archives/pulp-list/2013-February/msg00006.html

Only "clustering" solution I managed to came up with is to create a node
which will mirror all the repositories of master server. In case the master
fails I'll just redirect all the clients to the second server by configuration
management. I'd however have to also update all the repositories feeds
during the switch to point to right places.
Also usability of a load balancer in this scenario would be limited as the
secondary node would often lack the updates from master during sync.

As the active scenario didn't seem useful I tried to create a A/P setup.
First thing I tried was just rsync'ing the /var/lib/pulp contents to another
server set up as pulp server. When I started the http/qpidd on the passive
node I got no errors and the repositories seemed to work fine. Is this
approach OK or is there something I can break badly? It'd be nice to see
the cat living once I open the lid...

If there is anybody else with some different HA setup I'd really love to
hear about it.

// Mat

_______________________________________________
Pulp-list mailing list
Pulp-list at redhat.com
https://www.redhat.com/mailman/listinfo/pulp-list