[RFE] Support nokeepalive for serial console in the ipmitool driver

[Łukasz Leszczuk]

When serial console is started, there is no way to tell ipmitool to use `nokeepalive` option. In certain cases lack of this flag breaks long running sol sessions.

Use of this flag should be customizable by either driver_info or ironic.conf.

Reference:
- https://github.com/openstack/ironic/blob/bc330c934b848efb700ead6f5681ab9412340511/ironic/drivers/modules/ipmitool.py#L1207
- http://linux.die.net/man/1/ipmitool

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/372830

[OpenStack Infra (hudson-openstack)]

Related fix proposed to branch: master
Review: https://review.openstack.org/446571

[Galyna Zholtkevych (gzholtkevych)]

We cannot just change this option to nokeepalive. It is not good idea even to create this possibility.
The next solution is proposed:

1) using systemd init system. Logs using ironic-conductor as service unit:
http://paste.openstack.org/show/604822/

2) upon conductor failure stop consoles if ironic conductor exited due to some programming unexpected error (related patch above).

[Galyna Zholtkevych (gzholtkevych)]

Along with this change the patch https://review.openstack.org/372830 is welcome to be abandoned

[Galyna Zholtkevych (gzholtkevych)]

At the comment #3 first case covers second one, so just need to update console doc

[Madhuri Kumari (madhuri-rai07)]

Hi Galyna,

I tried to test this case with ironic-conductor managed by systemd and below is my observation:

Steps:
1. Started a console process by::

$ ironic node-set-console-mode node-0 true
$ ironic node-validate node-0
+------------+--------+---------------+
| Interface | Result | Reason |
+------------+--------+---------------+
| boot | True | |
| console | True | |
| deploy | True | |
| inspect | None | not supported |
| management | True | |
| network | True | |
| power | True | |
| raid | True | |
| storage | True | |
+------------+--------+---------------+
$ ironic node-get-console node-0
+-----------------+------------------------------------------------------------------+
| Property | Value |
+-----------------+------------------------------------------------------------------+
| console_enabled | True |
| console_info | {u'url': u'http://10.223.197.184:7777', u'type': u'shellinabox'} |
+-----------------+------------------------------------------------------------------+

2. The console process was started::

$ ps -eaf |grep shellinabox
ubuntu 75377 1 0 11:12 ? 00:00:00 shellinaboxd -t -p 7777 --background=/tmp/894fd6dc-b2c4-47bd-8054-0d46fce595cf.pid -s /:1000:1000:HOME:ipmitool -H 10.223.197.184 -I lanplus -U admin -f /tmp/894fd6dc-b2c4-47bd-8054-0d46fce595cf.pw -v sol activate
ubuntu 75378 75377 0 11:12 ? 00:00:00 shellinaboxd -t -p 7777 --background=/tmp/894fd6dc-b2c4-47bd-8054-0d46fce595cf.pid -s /:1000:1000:HOME:ipmitool -H 10.223.197.184 -I lanplus -U admin -f /tmp/894fd6dc-b2c4-47bd-8054-0d46fce595cf.pw -v sol activate
ubuntu 75533 74715 0 11:17 pts/1 00:00:00 grep --color=auto shellinabox

3. Stopped ironic-conductor service::
$ sudo systemctl ironic-conductor stop

4. Still the shellinaboxd processes were running.
5. Restarted ironic-conductor.
6. Disabled the node's console::
$ironic node-set-console-mode node-0 false

7. After this the console processes were successfully killed::
$ ps -eaf |grep shellinabox
ubuntu 77041 74715 0 12:11 pts/1 00:00:00 grep --color=auto shellinabox

$ ironic node-get-console node-0
+-----------------+-------+
| Property | Value |
+-----------------+-------+
| console_enabled | False |
| console_info | None |
+-----------------+-------+

CONCLUSION:
As per your comment #3, using default configuration of unit service file processes of same cgroup will be also killed. But the shellinabox process was running.

Can you please check this and let me know your opinion?

Thanks,
Madhuri

[OpenStack Infra (hudson-openstack)]

Change abandoned by Dmitry Tantsur (<email address hidden>) on branch: master
Review: https://review.openstack.org/372830
Reason: Abandoning due to inactivity

[OpenStack Infra (hudson-openstack)]

Change abandoned by Dmitry Tantsur (<email address hidden>) on branch: master
Review: https://review.openstack.org/446571
Reason: Abandoning due to inactivity