Open jelly opened 9 months ago
Yes please -- currently you have to specify a config like:
- name: set up kdump
hosts: localhost
vars:
kdump_target:
type: ssh
location: "admin@10.111.113.2"
kdump_ssh_server: "10.111.113.2"
kdump_ssh_user: "admin"
roles:
- linux-system-roles.kdump
... and apparently all of the keys location
, kdump_ssh_server
, kdump_ssh_user
are required. These are redundant and can contradict themselves.
We uncovered an issue when we don't provide a username and set the location to example.com
but do set kdump_ssh_server and kdump_ssh_user. This would lead to a kdump.conf with
-ssh admin@example.com
+ssh example.com
This breaks kdump.service, as ssh
is required to be user@example.com
.
See https://bugzilla.redhat.com/show_bug.cgi?id=2251216 for the kexec-tools side of this -- but I suspect they don't actually want to accept ssh hostname
, and it's just giving a bad error message. I think it'd be preferable in this case to only specify kdump_ssh_{server,user}
and not kdump_target.location
-- that will both fix the redundancy and also avoid that kexec-tools bug.
We have been working on generating an ansible role based on the current kdump configuration in Cockpit and noticed that with an ssh target
kdump_ssh_user
andkdump_ssh_server
while they could be found out from the kdump location.Without it the role fails.
I couldn't find any documentation on
kdump_ssh_user
andkdump_ssh_server
do the exist to copy the ssh key to the configured kdump user in the kdump location so that you could have a special kdump user to write the dump too?