the heat engine works slow compared with the direct provisioning engine
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Sahara |
Fix Released
|
High
|
Alexander Ignatov |
Bug Description
This bug was detected during integration tests execution. If we provision two equal clusters with direct engine and heat engine, the 'heat' cluster's state changes to "Active" 10-20 minutes later compared with the 'direct' cluster.
When we use the direct engine, we use Nova client to associate a floating IP address with an instance. But when we use the heat engine, Heat uses Neutron client to associate a floating IP address. Then we poll Nova to get the floating IP address and use it to configure a cluster. The problem is that Nova gets information about the floating IP address associated with the instance after 10-20 minutes if the IP was associated using the Neutron client.
For example, this is the console log of integration tests, and the only difference between them is the provisioning engine:
https:/
https:/
Changed in savanna: | |
assignee: | nobody → Alexander Ignatov (aignatov) |
Changed in savanna: | |
status: | Fix Committed → Fix Released |
Changed in sahara: | |
milestone: | icehouse-3 → 2014.1 |
Thanks Sergey! Bad to know that... /bugs.launchpad .net/tempest/ +bug/1262529.
I just wanted to mention that this bug is found in other projects as well https:/
Also we can crate a workaround in the code, to poll not nova VMs directly, but port list and grep floating ip in instance by its private ip, but we shoudl change the polling part in way to use neutron client instead of nova