libvirt users - Oct 2016 - Re: Fwd: Problems connecting to the libvirtd server

On 12.10.2016 16:06, Stefano Ricci wrote:
>>> ---------- Forwarded message ----------
>>> From: Stefano Ricci <sfn.rcc@gmail.com>
>>> Date: 2016-10-11 12:35 GMT+02:00
>>> Subject: Re: [libvirt-users] Problems connecting to the libvirtd
server
>>> To: Michal Privoznik <mprivozn@redhat.com>
>>>
>>
>> [Please don't top post on technical lists]
> Sorry I was wrong
> 
>>>
>>> There is not AppArmor and SElinux is disabled.
>>> The permits of the sock are as follows:
>>> libvirt-sock-admin 0700 (rwx) root: root
>>> libvirt-sock 0700 (rwx) root: root
>>> libvirt-sock-ro 0777 (rwxrwxrwx) root: root
>>> The libvirtd process runs with the root user.
>>
>> Okay, the permissions look okay.
> Good
> 
>>>
>>> Below you will find the link to the requested log:
>>> http://pastebin.com/0jtpWU0h (libvirt_client.log command
LIBVIRT_DEBUG
>>> = 1 virsh -c qemu: /// system list --all)
>>
>> And from the client logs it looks like client is trying to connect but
>> the daemon is not replying. So client just hangs waiting for server
>> reply. So I've looked into daemon logs and found this:
>>
>> 2016-10-10 12:10:39.138+0000: 1383: error : virPidFileAcquirePath:422 :
>> Failed to acquire pid file
'/var/lib/libvirt/qemu/capabilities.pidfile':
>> Risorsa temporaneamente non disponibile
>>
>> (with my little Italian knowledge, this is Resource temporary
unavailable)
>>
>> This suggests that there might be a different libvirt daemon running.
>> Can you confirm that?
> If I do a ps I am one libvirtd daemon and one of qemu and
> /var/lib/libvirt/qemu/capabilities.pidfile file exists
> The file contains the correct PID of the process qemu.
Okay, so what's the status of the qemu process? Because what might be
happening here is that qemu driver is still starting up, so it is not
accepting connections just yet. And because qemu process it was talking
to (whilst fetching the caps) hung up, the driver basically failed to
start (and is still in startup phase).

Michal

I checked with qmp and  your consideration was correct, the status of
qemu process is in prelaunch and running equals false.
But I do not understand why

Stefano

2016-10-12 10:43 GMT+02:00 Michal Privoznik
<mprivozn@redhat.com>:
> On 12.10.2016 16:06, Stefano Ricci wrote:
>>>> ---------- Forwarded message ----------
>>>> From: Stefano Ricci <sfn.rcc@gmail.com>
>>>> Date: 2016-10-11 12:35 GMT+02:00
>>>> Subject: Re: [libvirt-users] Problems connecting to the
libvirtd server
>>>> To: Michal Privoznik <mprivozn@redhat.com>
>>>>
>>>
>>> [Please don't top post on technical lists]
>> Sorry I was wrong
>>
>>>>
>>>> There is not AppArmor and SElinux is disabled.
>>>> The permits of the sock are as follows:
>>>> libvirt-sock-admin 0700 (rwx) root: root
>>>> libvirt-sock 0700 (rwx) root: root
>>>> libvirt-sock-ro 0777 (rwxrwxrwx) root: root
>>>> The libvirtd process runs with the root user.
>>>
>>> Okay, the permissions look okay.
>> Good
>>
>>>>
>>>> Below you will find the link to the requested log:
>>>> http://pastebin.com/0jtpWU0h (libvirt_client.log command
LIBVIRT_DEBUG
>>>> = 1 virsh -c qemu: /// system list --all)
>>>
>>> And from the client logs it looks like client is trying to connect
but
>>> the daemon is not replying. So client just hangs waiting for server
>>> reply. So I've looked into daemon logs and found this:
>>>
>>> 2016-10-10 12:10:39.138+0000: 1383: error :
virPidFileAcquirePath:422 :
>>> Failed to acquire pid file
'/var/lib/libvirt/qemu/capabilities.pidfile':
>>> Risorsa temporaneamente non disponibile
>>>
>>> (with my little Italian knowledge, this is Resource temporary
unavailable)
>>>
>>> This suggests that there might be a different libvirt daemon
running.
>>> Can you confirm that?
>> If I do a ps I am one libvirtd daemon and one of qemu and
>> /var/lib/libvirt/qemu/capabilities.pidfile file exists
>> The file contains the correct PID of the process qemu.
>
> Okay, so what's the status of the qemu process? Because what might be
> happening here is that qemu driver is still starting up, so it is not
> accepting connections just yet. And because qemu process it was talking
> to (whilst fetching the caps) hung up, the driver basically failed to
> start (and is still in startup phase).
>
> Michal

On 12.10.2016 19:16, Stefano Ricci wrote:
> I checked with qmp and  your consideration was correct, the status of
> qemu process is in prelaunch and running equals false.
> But I do not understand why
[Please don't top post on technical lists]

Well, I don't have any idea either. I mean, the daemon logs you provided
are from the second run of libvirtd. Maybe this is qemu or kernel bug?
If you kill the stalled process and then start libvirtd again, can you
connect then or does it still hang?

Michal