On Wed, Sep 19, 2018 at 10:23 AM Cody <codeology.lab(a)gmail.com&gt; wrote: > > Hello there, > > I have another question about scaling TripleO. > > According to the TripleO documentation for RHOPS 13/Queens [1], "you > cannot scale up or scale down a custom role that contains > OS::TripleO::Services::Pacemaker or > OS::TripleO::Services::PacemakerRemote services." > > Does that only apply to the role itself or the pcmk-managed services > within the role? For instance, if I already have 3 nodes running DB > and messaging services with a composable HA role, can I later create > another custom role just to scale up the messaging service again? > > Thank you very much! > > [1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/... > > Regards, > Cody > _______________________________________________ > users mailing list > users(a)lists.rdoproject.org > http://lists.rdoproject.org/mailman/listinfo/users > > To unsubscribe: users-unsubscribe(a)lists.rdoproject.org It applies to all services which are managed by Pacemaker, including the associated virtual IPs. The same services that can be separated from the Pacemaker services are the ones that can be moved to other roles and don't have the same scaling constraints. For instance, you could probably add X additional API worker nodes in a custom role, and they would be added as back-ends behind HAProxy, but you couldn't add any HAProxy nodes, since Pacemaker manages the virtual IPs that HAProxy binds to. Unfortunately, Pacemaker manages Rabbit MQ, so you are stuck with 3 nodes. -- Dan Sneddon | Senior Principal OpenStack Engineer dsneddon(a)redhat.com | redhat.com/openstack dsneddon:irc | @dxs:twitter

Hi Cody, On Thu, Sep 20, 2018 at 09:17:24AM -0400, Cody wrote: > Thank you, Dan. I really appreciate the answer. > > One last thing on the same subject. The RHOP 13 Director Installation > Manual says for the controller role that "environments with one node > can be used for testing purposes, environments with two nodes or more > than three nodes are not supported." [1] That effectively means either > one or three controller nodes are supported. To combine this with the > previous answer, could I reach the following understanding that: > > 1) As long as a role contains Pacemaker and its managed services, the > role is non-scalable after initial deployment; That is correct. Note that we merged code in Rocky so that this should actually start to work. We have only tested some specific scenarios (namely scaling one controller to three), but this should start working with rocky and feedback is welcome on that. > 2) For every Pacemaker managed service, the number of nodes is limited > to either one or three. That is correct, with one caveat, we generally support an odd number of cluster nodes in total, for quorum reasons. hth, Michele > Is that right? > > [1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/... > > Regards, > Cody > On Wed, Sep 19, 2018 at 8:46 PM Dan Sneddon <dsneddon(a)redhat.com&gt; wrote: > > > > > > > > On Wed, Sep 19, 2018 at 10:23 AM Cody <codeology.lab(a)gmail.com&gt; wrote: > >> > >> Hello there, > >> > >> I have another question about scaling TripleO. > >> > >> According to the TripleO documentation for RHOPS 13/Queens [1], "you > >> cannot scale up or scale down a custom role that contains > >> OS::TripleO::Services::Pacemaker or > >> OS::TripleO::Services::PacemakerRemote services." > >> > >> Does that only apply to the role itself or the pcmk-managed services > >> within the role? For instance, if I already have 3 nodes running DB > >> and messaging services with a composable HA role, can I later create > >> another custom role just to scale up the messaging service again? > >> > >> Thank you very much! > >> > >> [1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/... > >> > >> Regards, > >> Cody > >> _______________________________________________ > >> users mailing list > >> users(a)lists.rdoproject.org > >> http://lists.rdoproject.org/mailman/listinfo/users > >> > >> To unsubscribe: users-unsubscribe(a)lists.rdoproject.org > > > > > > It applies to all services which are managed by Pacemaker, including the associated virtual IPs. The same services that can be separated from the Pacemaker services are the ones that can be moved to other roles and don't have the same scaling constraints. For instance, you could probably add X additional API worker nodes in a custom role, and they would be added as back-ends behind HAProxy, but you couldn't add any HAProxy nodes, since Pacemaker manages the virtual IPs that HAProxy binds to. Unfortunately, Pacemaker manages Rabbit MQ, so you are stuck with 3 nodes. > > > > -- > > Dan Sneddon | Senior Principal OpenStack Engineer > > dsneddon(a)redhat.com | redhat.com/openstack > > dsneddon:irc | @dxs:twitter > -- > users mailing list > users(a)lists.rdoproject.org > http://lists.rdoproject.org/mailman/listinfo/users > > To unsubscribe: users-unsubscribe(a)lists.rdoproject.org > -- Michele Baldessari <michele(a)acksyn.org&gt; C2A5 9DA3 9961 4FFB E01B D0BC DDD4 DCCB 7515 5C6D