octavia test_basic_http_traffic on c9 times out as it asks for password while connecting to the vm
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
Critical
|
Unassigned |
Bug Description
octavia_
This should not be the proper behavior. On tempest tests, the ssh should be injected in the vm, and this should be accessed without the user interact.
The test is hanging exactly because of this, if you run the tempest manually, at some point, after the vm is on ACTIVE status, and the test tries to ssh to the vm, it ask for the password.
We tried to add into tempest.conf the image_ssh_password to see if we can workaround this, but no success.
We are pretty sure this is related to octavia test itself, because we teste the test_network_
Bellow is the tempest.conf section for octavia that is being used:
```
[load_balancer]
test_server_path = "/usr/lib/
test_with_ipv6 = False
enable_
admin_role = admin
RBAC_test_type = owner_or_admin
member_role = member
region = regionOne
enabled_
```
Right now, we are not sure if this requires other configuration on load_balancer section or in the code of the test itself.
We were able to access the cirros VM manually, and we can confirm that the authorized_key is there on .ssh/authorized
We also notice adding a different ssh key in the authorized_keys did not work. so it seems to be how the vm is being created (again, the basic_network_
Looks like a problem with how the VM boots up in octavia.
Changed in tripleo: | |
status: | New → Triaged |
importance: | Undecided → Critical |
milestone: | none → yoga-2 |
tags: | added: promotion-blocker |
Changed in tripleo: | |
status: | Triaged → Fix Released |
There's an incompatibility between Centos 9 stream ssh client and cirros ssh server, the scp connection initiated in the octavia- tempest- plugin fails (but the RemoteClient class works).
A workaround would be to extend the PubkeyAcceptedK eyTypes for scp with:
-o PubkeyAcceptedK eyTypes= +rsa-sha2- 256,rsa- sha2-512