[ovirt-users] Problem with hosted engine setup - vsdmd does not start (RESOLVED)

[Yedidyah Bar David]

On Thu, Dec 17, 2015 at 6:45 PM, Will Dennis <wdennis at nec-labs.com> wrote:
> The vdsmd startup failing problem is now resolved :) The root cause was that the /etc/sudoers file (that we maintain a customized version of which thru Ansible) was missing the "#includedir /etc/sudoers.d" stanza, which caused a sudo problem with the vdsmd startup, since the sudoers include file "50-vdsm" wasn't being parsed.
>
> In going back and reviewing the thread, this was the first post that pointed out the real problem...
>
> -----Original Message-----
> From: Fabian Deutsch [mailto:fdeutsch at redhat.com]
> Sent: Tuesday, December 01, 2015 12:58 AM
> To: Will Dennis
> Cc: Simone Tiraboschi; users
> Subject: Re: [ovirt-users] Problem with hosted engine setup - vsdmd does not start
>
> On Tue, Dec 1, 2015 at 4:52 AM, Will Dennis <wdennis at nec-labs.com> wrote:
>> Any clues out of the strace of vdsm?
>
> read(9, "sudo: a password is required\n", 4096) = 29
>
> Could it be that sudo is not configured to operate passwordless?
>
> The strat-up can then fail, because sudo requires a ty, but this isn't available during service start.
>
> - fabian
>
>> On Nov 25, 2015, at 11:29 AM, Simone Tiraboschi <stirabos at redhat.com> wrote:
>>
>>> On Wed, Nov 25, 2015 at 5:22 PM, Willard Dennis <wdennis at nec-labs.com> wrote:
>>> [root at ovirt-node-01 ~]# sudo -u vdsm /bin/bash
>>> bash-4.2$ /usr/share/vdsm/vdsm
>>> bash-4.2$ echo $?
>>> 1
>>
>> Can you please use strace on it?
>
> -----Message End-----
>
> However, not knowing about the missing sudoers include file, we continued to chase a vdsm config problem (uninstall / reinstall vdsm RPM pkgs, & use "vdsm-tool configure --force")
>
> Shame on me, it looks like I didn't think to check /var/log/messages for "vdsm"-string entries until today, where we found the "vdsm user could not manage to run sudo operation: (stderr: ['sudo: sorry, you must have a tty to run sudo']). Verify sudoer rules configuration” message that ultimately led to the solution :(
>
> However, not sure if it's possible to do so, but could /usr/share/vdsm/vdsm be made to emit a error to STDERR if it cannot run correctly? That's the only thing I could think of that may have helped me to resolve this earlier with what I did do...

Well, I can think of other things as well, such as doing such a test
during rpm/yum install stage, etc., but the question is how far should
we go - there are many other ways to break ovirt, and we can't
possibly consider all of them. That said, I guess patches are welcome
(but you'll need to discuss this with a vdsm maintainer, not me,
probably on gerrit and not here).

>
> Thanks for everyone's kind assistance, this really is a great community! :)

Thanks a lot for the time spent and the report!

Best,
-- 
Didi