[library] Deployment fails: DNS resolving is broken on nodes with Mongo role
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Fix Committed
|
High
|
Oleksiy Molchanov |
Bug Description
Fuel version info (6.1 build # 234): http://
Environment deployment fails on Ubuntu if 'Ceilometer' is enabled, because after deployment (dns-client.pp) of nodes with 'mongo' role, they can't resolve DNS names:
http://
Steps to reproduce:
1. Create new environment: Ubuntu, Neutron+Vlan, Ceph for all, Ceilometer
2. Add 3 controller+mongo nodes and 1+ compute+ceph nodes
3. Enable 'Classic' provision method on settings tab
4. Start deployment
Expected result:
- cluster successfully deployed and passes health checks
Actual:
- cluster deployment failed
Here is a content of '/etc/resolv.conf/' file on controller+mongo node after deployment failed:
http://
and tasks deployment sequence:
http://
As you can see node was configured to use management VIP as DNS resolver before controller roles were deployed.
Changed in fuel: | |
assignee: | Fuel Library Team (fuel-library) → Oleksiy Molchanov (omolchanov) |
Changed in fuel: | |
status: | New → In Progress |
Seems that image based provisioning on Ubuntu isn't affected, because resolvconf is unable to update '/etc/resolf.conf':
root@node-1:~# cat /etc/resolv.conf /update. d/libc: Warning: /etc/resolv.conf is not a symbolic link to /run/resolvconf /resolv. conf
# re-generated by cloud-init boothook only at the first boot;
search mirantis.com
domain mirantis.com
nameserver 172.16.39.130
root@node-1:~# resolvconf -u
/etc/resolvconf
root@node-1:~# ls -1l /etc/resolv.conf
-rw-r--r-- 1 root root 127 Mar 31 11:54 /etc/resolv.conf
but I think this breaks 'external dns' feature (see blueprint external- dns-ntp- support)