[ovirt-devel] [ OST Failure Report ] [ oVirt Master ] [ bootstrap.verify_add_hosts ] [ 18/10/17 ]

[Martin Perina]

So the real issue on adding a host is the same as I've today described in
[2] and most probably caused by [3] (I reverted engine in my dev env prior
this patch and host deploy finished successfully).

Allon, do you have time to post a fix? If not I'll try to dig into your
change and related networking code to post it ...


[2] https://bugzilla.redhat.com/show_bug.cgi?id=1504005
[3] https://gerrit.ovirt.org/#/c/82545/

On Thu, Oct 19, 2017 at 11:12 AM, Martin Perina <mperina at redhat.com> wrote:

>
>
> On Thu, Oct 19, 2017 at 11:04 AM, Martin Perina <mperina at redhat.com>
> wrote:
>
>>
>>
>> On Thu, Oct 19, 2017 at 10:58 AM, Dan Kenigsberg <danken at redhat.com>
>> wrote:
>>
>>> On Thu, Oct 19, 2017 at 10:29 AM, Martin Perina <mperina at redhat.com>
>>> wrote:
>>> >
>>> >
>>> > On Thu, Oct 19, 2017 at 7:35 AM, Dan Kenigsberg <danken at redhat.com>
>>> wrote:
>>> >>
>>> >> On Wed, Oct 18, 2017 at 2:40 PM, Daniel Belenky <dbelenky at redhat.com>
>>> >> wrote:
>>> >>>
>>> >>> Hi all,
>>> >>>
>>> >>> The following test is failing: 002_bootstrap.verify_add_hosts
>>> >>> All logs from failing job
>>> >>> Only 2 engine patches participated in the test, so the suspected
>>> patches
>>> >>> are:
>>> >>>
>>> >>> https://gerrit.ovirt.org/#/c/82542/2
>>> >>> https://gerrit.ovirt.org/#/c/82545/3
>>> >>>
>>> >>> Due to the fact that when this error first introduced we had another
>>> >>> error, the CI can't automatically detect the specific patch.
>>> >>>
>>> >>> Error snippet from logs: ovirt-host-deploy-ansible log (Full log)
>>> >>>
>>> >>> TASK [ovirt-host-deploy-firewalld : Enable firewalld rules]
>>> >>> ********************
>>> >>> failed: [lago-basic-suite-master-host-0] (item={u'service':
>>> >>> u'glusterfs'}) => {"changed": false, "failed": true, "item":
>>> {"service":
>>> >>> "glusterfs"}, "msg": "ERROR: Exception caught:
>>> >>> org.fedoraproject.FirewallD1.Exception: INVALID_SERVICE:
>>> 'glusterfs' not
>>> >>> among existing services Permanent and Non-Permanent(immediate)
>>> operation,
>>> >>> Services are defined by port/tcp relationship and named as they are
>>> in
>>> >>> /etc/services (on most systems)"}
>>> >>>
>>> >>>
>>> >>> Error from HOST 0 firewalld log:
>>> >>> lago-basic-suite-master-host-0/_var_log/firewalld/ (Full log)
>>> >>>
>>> >>> 2017-10-15 16:51:24 ERROR: INVALID_SERVICE: 'glusterfs' not among
>>> >>> existing services
>>> >>
>>> >>
>>> >> Ondra, would such an error propagate through the playbook to Engine
>>> and
>>> >> fail the add-host flow? (I think it should!)
>>> >
>>> >
>>> > We didn't do that so far, because of EL 7.3
>>> > . We need firewalld from 7.4 to have all available services in place (I
>>> > don't remember but I think imageio service was the one delivered only
>>> in
>>> > firewalld from 7.4). So  up until now we ingore non-existent firewalld
>>> > service, but if needed we can turn this on and fail host deploy.
>>>
>>> Ok, so for now your "luckily" off the hook and not the reason of failure.
>>>
>>> >>
>>> >>
>>> >> Do you know which package provide the glusterfs firewalld service,
>>> and why
>>> >> it is missing from the host?
>>> >
>>> >
>>> > So we have used 'glusterfs' firewalld service per Sahina
>>> recommendation,
>>> > which is included in glusterfs-server package from version 3.7.6 [1].
>>> But
>>> > this package is not installed when installing packages for cluster with
>>> > gluster capabilities enabled. So now I'm confused: don't we need
>>> > glusterfs-server package? If not and we need those ports open because
>>> they
>>> > are used by services from different already installed glusterfs
>>> packages,
>>> > shouldn't the firewalld configuration be moved from glusterfs-server to
>>> > glusterfs package?
>>>
>>> glusterfs-cli.rpm is required to consume gluster storage (virt use
>>> case), but I don't recall that it needs open ports.
>>>
>>
>> ​It was there even for IPTables, if gluster support is enabled on
>> cluster, then gluster specific ports were enabled even with IPTables.
>> FirewallD feature continues to use that.
>> ​
>>
>>
>>> glusterfs-server.rpm is required to provide gluster storage (gluster use
>>> case).
>>> If I recall correctly, firewalld feature has differentiated between
>>> the two; opening needed ports only when relevant.
>>>
>>
>> ​Right, but if gluster services are configured for firewalld it means
>> that the host has been added to the cluster with gluster feature enabled
>> and not only virt
>> ​
>>
>>
>>>
>>> >
>>> >
>>> > [1] https://bugzilla.redhat.com/show_bug.cgi?id=1057295
>>> >
>>> >
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/devel/attachments/20171019/ce4c0c58/attachment-0001.html>