Re: wiki: How to fix failed autotest clients

[James Laska]

On Fri, 2010-06-25 at 08:28 -0400, Kamil Paral wrote:
>  I've been reading through:
>  
>    https://fedoraproject.org/wiki/How_to_fix_failed_autotest_clients
>  
>  There is this example:
>  $ atest host mod -y my-failed-system.example.com
>  
>  The problem is option -y is no longer available in atest command.
>  I haven't found any other way how to mark failed system as ready.
>  I have also not found any way how to start verification of that system.
>  The only possible way I know about is through the web interface, 
>  there's a button for verification.
>  
>  Am I missing something? I would like to correct the docs, but
>  I fail to find out how. 
That doc is pretty weak, apologies.  I wanted to provide some common
steps for debugging why the system is not responding ... and couldn't
figure out how to confine the page to just autotest ... and not a
general guide for debugging linux connectivity issues.  Any tips/ideas
appreciated.

Looks like the 'atest host mod -y' option is no longer available with
autotest-0.12.0.  I don't seen an obvious way to schedule a 'verify' job
against a host.  

Upon further inspection, the autotest scheduler daemon (monitor_db.py
and it's babysitter monitor_db_babysitter) has an option to
--recover-hosts.  We may wish to enable this in our autotestd SysV
initscript ... and it will attempt to recover any test clients that had
connectivity issues.

This leaves the task to the autotest admin to determine the failure
cause and bring the test client back online.

Thanks,
James