Fuel 8.0 puppet manifest fails on second run if node hostname capitalized

[Max Lvov]

Fuel 8.0 3 Controllers, 5 Computes, 4 Ceph-osd, 3 Telemetry

How to reproduce:
1 Have deployment on primary controller failed, but after mysql users created
2 Without stopping deployment and resetting env, wait until env become failed
3 Push deploy

What I expect:
Deployment will started again and will reach ta least the same point of puppet manifest

What I get:
ERROR 1396 (HY000): Operation CREATE USER failed for 'sahara'@'rspc1ctrlr1'
because it already exists
http://bugs.mysql.com/bug.php?id=28331

[Sergii Golovatiuk (sgolovatiuk)]

not valid for 10.0 and 9.0

[Max Lvov (usrleon)]

Run into this issue again with same conditions, after primary controller failed.

[Max Lvov (usrleon)]

Found root cause, this issue appears when we use capitalized host names and node names.

Fix:
Change this line https://github.com/openstack/fuel-library/blob/stable/8.0/deployment/puppet/osnailyfacter/modular/sahara/db.pp#L23
Add downcase:
$allowed_hosts = [ downcase($node_name), 'localhost', '127.0.0.1', '%' ]

Please retest for 9 and 10 with capitalized hostnames.

[Max Lvov (usrleon)]

Also, I'm getting now errors related to networks creation. Log attached.

[Max Lvov (usrleon)]

Last error related to networks caused by https://bugs.launchpad.net/fuel/+bug/1471568

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/8.0
Review: https://review.openstack.org/389084

[Rodion Tikunov (rtikunov)]

Steps to reproduce:
1) echo "include blabla" >> /etc/puppet/modules/osnailyfacter/modular/sahara/sahara.pp
2) fuel rel --sync-deployment-tasks --dir /etc/puppet
3) activate Sahara, Murano, Ceilometer options in cluster settings
4) create cluster 1 controller+mongo and 2 compute nodes
5) deploy cluster
6) cluster deployment will fail because of error "module blabla doesn't exists"
7) delete "include blabla" from /etc/puppet/modules/osnailyfacter/modular/sahara/sahara.pp
8) fuel rel --sync-deployment-tasks --dir /etc/puppet
9) deploy cluster

[Alexey Stupnikov (astupnikov)]

A steps to reproduce from message #7 didn't worked for me on MOS7 and MOS8 with latest MUs installed. I was able to successfully deploy failed cluster. I have doublechecked if db.pp was applied before failure and it was, so I think that this bug should go to Invalid for 7.0-updates and should go to Inclomplete for 8.0-updates.

@usrleon can you please check if latest MU fixes this problem on MOS8 and move bug back to work if it is not?

[Max Lvov (usrleon)]

@astupnikov of course I can't. This deployment was about half of year ago and I published all possible information. Even though you can't reproduce look at code, it's obviously the case.

[Alexey Stupnikov (astupnikov)]

@usrleon I agree that if we apply MySQL ACLs with allowed host list like in [1] twice we will reproduce reported MySQL bug. But since you have reported this problem we have updated MySQL, improved deployment process and updated some other packages. As a result, I think that it is ok if we do not get into this issue anymore. I think that we shouldn't patch something if the problem is not there.

BTW, I can provide you 8.0 env or try other steps to reproduce you suggest if you want to continue investigation.

[1] https://github.com/openstack/fuel-library/blob/stable/8.0/deployment/puppet/osnailyfacter/modular/sahara/db.pp#L23

[Alexey Stupnikov (astupnikov)]

Closing as Invalid as we are unable to reproduce this issue with latest MU installed.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Andreas Jaeger (<email address hidden>) on branch: stable/8.0
Review: https://review.opendev.org/389084
Reason: This repo is retired now, no further work will get merged.