MongoDB Can't find master host for replicaset tripleo.

[Jiří Stránský]

The deployment sometimes fails with:

Error: /Stage[main]/Tripleo::Profile::Base::Database::Mongodb/Mongodb_replset[tripleo]: Could not evaluate: Can't find master host for replicaset tripleo.

This happens in Step 3 of controller deployment:

| AllNodesDeploySteps | c2be42a8-3108-48e6-a416-244a557c7195 | OS::TripleO::PostDeploySteps | CREATE_FAILED | 2016-09-16T13:19:43Z | overcloud |
| ControllerDeployment_Step3 | e6adf814-fef2-444c-9123-d798a05e3115 | OS::Heat::StructuredDeploymentGroup | CREATE_FAILED | 2016-09-16T13:24:32Z | overcloud-AllNodesDeploySteps-lflgrlcmum7v |
| 0 | a7ff06ab-f548-4d48-aeb6-3c9ecf6f54c9 | OS::Heat::StructuredDeployment | CREATE_FAILED | 2016-09-16T13:29:04Z | overcloud-AllNodesDeploySteps-lflgrlcmum7v-ControllerDeployment_Step3-wks6g5uos4gt |

This failure happened for me locally and also in CI job here:

http://logs.openstack.org/96/365796/14/check-tripleo/gate-tripleo-ci-centos-7-ovb-ha/8d79c0f/logs/overcloud-controller-0/var/log/messages

Interesting is that it seems to succeed creating the replset at step 2 but then fails to "reassure" the resource at step 3.

[Jiří Stránský (jistr)]

https://review.openstack.org/#/c/371596/

[Jiří Stránský (jistr)]

Another job failed on this:

http://logs.openstack.org/87/370687/2/check-tripleo/gate-tripleo-ci-centos-7-ovb-ha/cbdb8fb/console.html

[Jiří Stránský (jistr)]

Tagging with CI as it causes intermittent CI failures.

[Tom Barron (tpb)]

I'm hitting this, and related mongodb repl set failures, with fresh packages, tripleo-ci instack and overcloud deploy today. Here's an example with the message cited in the bug header:

Error: /Stage[main]/Tripleo::Profile::Base::Database::Mongodb/Mongodb_replset[tripleo]: Could not evaluate: Can't find master host for replicaset tripleo

See http://paste.fedoraproject.org/429100/05324614/

Observe that ab bit before the Error msg there is the following Warning (from line 26 of the pastebin cited):

Warning: Scope(Class[Mongodb::Server]): Replset specified, but no replset_members or replset_config provided.

Naively, if there aren't replset_members then that could explain not beting able to find a master host for the replicaset.