Showing headlines posted by dba477

Perform two node cluster deployment Controller + Network&Compute (ML2&OVS&VXLAN). Another configuration available via packstack Controller+Storage+Compute&Network.

Following bellow is set of directives allowing to switch Openstack RDO Mitaka to using flat (vlan) external network provider,what allows to work with several external networks via single L3 router. Conversion supposed to be done doesn't depend in any way of how tenants are segregated either VLAN tagged networks or VXLAN (GRE) tunneling.

Python code posted down-here utilizes NeutronClient V2 API. It creates neutron router for particular tenant, interface to private sub-net and gateway to external flat network. As usual posting was inspired by question been posted at ask.openstack.org. Code is intentionally simplified to become understandable by people without strong python development background.

Quoting Official delorean documentation "The RDO project has a continuous integration pipeline that consists of multiple jobs that deploy and test OpenStack as accomplished by different installers.This vast test coverage attempts to ensure that there are no known issues either in packaging, in code or in the installers themselves .. "

Verification been done bellow is actually targeting conversion of HAProxy/Keepalived (Active/Active) 3 Node Controller which design was suggested for RDO Liberty in https://github.com/beekhof/osp-ha-deploy/blob/master/HA-keep... to be able support Compute Nodes running in DVR mode. The core issue on Liberty was resolved for Mitaka , see upstream record [RFE] Unable to create a router that's both HA and distributed

Just a reminder in Juno and Kilo DVR was available for deployments using VXLAN tunneling and required l2population activation on all nodes. One of new features of Liberty is DVR compatibility with ML2&OVS&VLAN deployed landscapes. On RDO Liberty packstack doesn't play so nicely doing VLAN deployment as in case of VXLAN tunneling

Current post follows up "Hackery to get going RDO Kilo on Fedora 23" To complete packstack run for two nodes Controller/Network and Compute setup I had to apply as pre-installation following patches, otherwise neutron puppet crashed on Fedora 23 . . .

Sequence of hacks required to get going RDO Kilo on Fedora 23 is caused by existence of many areas where Nova has a hard dependency on Glance v1. According to Launchpad Bug 1476770 I can see that for python-glanceclient status is "In Progress". So, to get things working right now only one option is available (for myself of course) . . . .

Post bellow presumes that your testing Swift storage is on VM emulating Storage Server (192.169.142.157) which has 3 devices attached /dev/vdb,/dev/vdc,/dev/vdd what allows to place in answer file entries like bellow. In case you need not just a POC, but real HA Swift deployment install RDO with CONFIG_INSTALL_SWIFT=n and make manual deployment on your management network - one Swift proxy node and at least 3 Swift Storage nodes.

Post below presumes that your testing Swift storage is located somewhere on workstation (say /dev/sdb1) is about 25 GB (XFS) and before running packstack (AIO mode for testing) following steps have been done . . .

Post below addresses several questions been posted at ask.openstack.org In particular, code below doesn't require volume UUID to be hard coded to start server attached to boot able cinder's LVM, created via glance image, which is supposed to be passed to script via command line. In the same way name of cinder volume and instance name may be passed to script via CLI.

This post briefly describes how to rebuild currently used openstack-packstack packages on CentOS 7.2 (RDO Liberty) and put in work patch proposed by bugzilla record "Bug 1254389 - Can no longer run packstack to maintain cluster" right away, i.e. avoiding waiting official procedure of pushing new openstack-packstack packages to stable repos.

I have noticed several questions (ask.openstack.org,stackoverflow.com ) regarding mentioned ongoing issue with mongodb-server and nagios when installing RDO Kilo 2015.1.1 on CentOS 7.2 via packstack. At the moment I see a hack provided bellow which might be applied as pre-installation step or a fix after initial packstack crash.

RDO deployment bellow has been done via straightforward RDO Kilo packstack run demonstrates that Storage Node might work as traditional iSCSI Target Server and each Compute Node is actually iSCSI initiator client. Notice that Glance,Cinder,Swift Services are not running on Controller.

Post is actually step by step procedure of creating DVR system working with two external networks been built via flat network provider. Question which several times was raised up at ask.openstack.org, however was not addressed properly.

Test bellow is targeting two potential problems :- 1. Creating HAProxyKeepalived 3 Node Controller in RDO Mitaka with router supporting VRRP && DVR at a time per https://github.com/beekhof/osp-ha-deploy/blob/master/HA-keep... in regards of using DVR on Compute nodes. 2.Creating DVR system working with two flat external networks.

Post bellow is addressing the question when AIO RDO Liberty Node has to have external networks of VLAN type with predefined vlan tags. Straight forward packstack --allinone install doesn't allow to achieve desired network configuration. External network provider of vlan type appears to be required

It is actually an update for Neutron on RDO Liberty of original blog entry http://blog.aaronorosen.com/implementing-high-availability-i... . I only attempted to make post understandable for people with no knowledge how to access cloud VMs having just private IPs and highlighted Neutron Commands which are not commonly known and provide option to create floating IP working as VIP providing HA for a couple on Ubuntu 14.04 cloud instances

The correct name of this post is supposed to be "Nova and Neutron workflow && CLI for HAProxy/Keepalived 3 Node Controller RDO Liberty in an appropriate amount of detail". It follows up http://lxer.com/module/newswire/view/222164/index.html . Neutron work-flow on Controller is described including OVS flow rules on external brigge created by flat external network provider and eth0 external interface as VLAN OVS port of bridge br-eth0

As far as to my knowledge Cisco's schema has been implemented :- Keepalived, HAProxy,Galera for MySQL Manual install, at least 3 controller nodes. I just highlighted several steps which as I believe allowed me to bring this work to success.