Ask Your Question

kolla-ansible deploy fails on DB create/connect (which works manually from within containers)

asked 2018-04-05 11:46:26 -0500

theque42 gravatar image

updated 2018-04-18 03:20:57 -0500

Still centos7, updated, with ansible 2.5, updated pip, and kolla-ansible from PIP. I get the same error almost every time, regardless of the number of controller nodes I deploy to(1 or 2) Last attempt had 2 nodes of each function. All nodes are always created using qcow-backing images from the same centos-base.

During deploy, the /neutron/tasks/bootstrap.yml, when trying to create the neutron database, it fails with:

"unable to find /var/lib/ansible/.my.cnf. Exception message: (2003, 'Can\\'t connect to MySQL server on \\'\\' (110 \"Connection timed out\")')"}

Where ever I look, there is no ".my.cnf" on any host (ctrl, neutron, etc) so should that's ONE problem I dont know how to solve, if it even should be available after the failed playbook.

The weirder problem is that I have NO problems connecting to .111 (the VIP address) if I manually enter "mysql -u haproxy -h <VIP> in either the maradb container, or kolla-toolbox container on either of the control nodes.

So HOW THE HECK can I troubleshoot this? HELP :'(

edit retag flag offensive close merge delete


connection to the database cluster is also made through the VIP

Eduardo Gonzalez gravatar imageEduardo Gonzalez ( 2018-04-06 02:13:03 -0500 )edit

So how do I figure out why it fails? Looking at both haproxy containers on my two neutron nodes, I can see established both local and vip address available, LISTENing sockets on VIP, and ESTABLISHED connections from node to both ctrl:er nodes, to maria db.

theque42 gravatar imagetheque42 ( 2018-04-07 05:09:30 -0500 )edit

2 answers

Sort by ยป oldest newest most voted

answered 2018-04-19 14:49:19 -0500

theque42 gravatar image

Here is at least part of the answer it seems....

edit flag offensive delete link more

answered 2018-05-04 10:29:05 -0500

flashydave gravatar image

updated 2018-05-04 10:29:56 -0500

I still consistently get the issue on a reconfigure or even a destroy/deploy using latest version and the additions in the commit in the other answer are present in my bootstrap.yml. This is on ubuntu 16.04 on the same server as the deployment. The mariadb container is running fine.

I have also tried using localhost or the server name explicitly (there was another page that hinted there might be an issue with unix sockets being in an unexpected location. However I have deployed OK with an earlier version of kolla-ansible

My hardware is not as fast as I would like and when doing a koll-ansible deploy with -vvv option piped out to a local file I did manage to get it successful just once but I cannot repeat it. Is there a timing dependency??

edit flag offensive delete link more


All my nodes are actually virtualized, so yes, I too have started thinking about a possible timing issue here. I gave up trying to make it work on "released" queens installed via pip, and I am now running kolla-ansible from github, which works better.

theque42 gravatar imagetheque42 ( 2018-05-05 13:31:06 -0500 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2018-04-05 11:46:26 -0500

Seen: 1,542 times

Last updated: May 04 '18