[Bug 1449682] Re: incorrect mysql credentials when deploying cinder, keystone HA against percona-cluster HA

[James Page]

As this bug is from last year, I'm really hoping this is not an issue;
however if it is please could you confirm the following:


a) which version of juju you are using; I suspect some sort of state consistency issue with regards to storage of shared-db passwords on the peer relation, prior to the introduction of leader election and storage in 1.25.x - if this env is pre 1.25.x, then this might be the cause (peer relation storage tended to be buggy)

b) If this was 1.25.x, then the values from leader storage should tell
us some more:

juju run --unit mysql/0 "leader-get"

as this is what is actually used for the canonical source of passwords
for leader-election enabled juju environments.

Marking 'Incomplete'; please set back to New if you're still seeing this
issue.

This might also be related to bug 1515976 as your openstack services are
cluster.

** Changed in: percona-cluster (Juju Charms Collection)
       Status: New => Incomplete

** Changed in: keystone (Juju Charms Collection)
       Status: New => Incomplete

** Changed in: cinder (Juju Charms Collection)
       Status: New => Incomplete

** Changed in: cinder (Juju Charms Collection)
   Importance: Undecided => Low

** Changed in: keystone (Juju Charms Collection)
   Importance: Undecided => Low

** Changed in: percona-cluster (Juju Charms Collection)
   Importance: Undecided => Low

-- 
You received this bug notification because you are a member of Ubuntu
OpenStack, which is subscribed to cinder in Juju Charms Collection.
Matching subscriptions: charm-bugs
https://bugs.launchpad.net/bugs/1449682

Title:
  incorrect mysql credentials when deploying cinder,keystone HA against
  percona-cluster HA

Status in cinder package in Juju Charms Collection:
  Incomplete
Status in heat package in Juju Charms Collection:
  Invalid
Status in keystone package in Juju Charms Collection:
  Incomplete
Status in percona-cluster package in Juju Charms Collection:
  Incomplete

Bug description:
  Context:
  *  openstack HA deployment using 1504 charms release
  * 'mysql' juju service deployed using percona-cluster charm r57
  * 'cinder' service deployed using cinder charm r91
  * staged deployment as:

  #1 deploy cinder num_units:3, mysql num_units:2
   (among others), with their hacluster relations
  #2 relate all OS services (ie cinder<->mysql, etc)
  #3 add 3rd unit to mysql

  FYI using mojo with
  lp:~jjo/openstack-mojo-specs/mojo-openstack-specs-add-mojo-bootstack-specs :
  - specs/bootstack/ha_phased/icehouse/manifest
  - specs/bootstack/ha_phased/icehouse/bootstack-example.yaml

  Issue:
  * after #3, cinder/0 fails with
  "CRITICAL cinder [-] OperationalError: (OperationalError) (1045, "Access denied for user 'cinder'@'172.20.0.184' (using password: YES)") None None"
  - http://paste.ubuntu.com/10928144/

  * cinder/0 (172.20.0.184) has a different password at mysql/0 than the rest:
  - http://paste.ubuntu.com/10928182/

  * cinder/0 shows different relation data against mysql/0,1 (/2 didn't finish deploying)
  - http://paste.ubuntu.com/10928280/

  * cinder/1,2 are stuck re-running shared-db-relation-changed hook
  (more than ~1d already), FYI number of this hook's lines at log:
  $ for i in cinder/{0..2};do juju ssh $i 'sudo grep -cH shared-db-relation-changed /var/log/juju/unit-cinder-?.log';done
  - http://paste.ubuntu.com/10928334/

To manage notifications about this bug go to:
https://bugs.launchpad.net/charms/+source/cinder/+bug/1449682/+subscriptions