[openstack-dev][cinder] question on cinder-volume A/A configuration

Gorka Eguileor Gimeno geguileo at redhat.com
Fri Sep 20 10:56:25 UTC 2019


On Thu, Sep 19, 2019 at 12:19 PM Chen CH Ji <jichenjc at cn.ibm.com> wrote:

> Hi stackers
>               I have a 1 controller + 2 compute nodes settings, same
> configuration works fine on Ocata version but failed to run on Stein
> version , I am using
>
> compute node1 and node 2 both use this backend and I can see only 1
> compute services
>
>
>  [root at controller ~]# cinder service-list
>
> +------------------+----------------+------+---------+-------+----------------------------+-----------------+
> | Binary           | Host           | Zone | Status  | State | Updated_at
>                 | Disabled Reason |
>
> +------------------+----------------+------+---------+-------+----------------------------+-----------------+
> | cinder-scheduler | controller     | nova | enabled | up    |
> 2019-09-19T09:16:21.000000 | -               |
> | cinder-volume    | FC at POWERMAX_FC | nova | enabled | up    |
> 2019-09-19T09:16:30.000000 | -               |
>
> +------------------+----------------+------+---------+-------+----------------------------+-----------------+
>

Hi,

The subject mentions cinder-volume A/A, but then you say you only have 1
controller node and there's only 1 scheduler and API service, so how is
it running as A/A?


> cinder.conf
>
> [backend]
> backend_host=FC at POWERMAX_FC
>
>
> and now I am creating 5 instances from nova at same time (boot from
> volume) , the scheduler will report those error time to time like following
> ,but actually the 2 services on both 2 compute nodes runs fine ..
>
> 2019-09-19 17:53:10.951 20916 WARNING cinder.scheduler.host_manager
> [req-19e722e8-1523-4121-8987-3cb450a8038e 071294a19fa8463788822565e0927fce
> f43175c07dc8415899d6b350dbede772 - default default] volume service is down.
> (host: FC at POWERMAX_FC)
>

You should check the cinder-volume logs to see the frequency of the
stats reports, since slow stats reports will make the scheduler think
the nodes are down.

If you had multiple controller nodes it could be out of sync clocks.

Cheers,
Gorka.


>
> those settings works fine on O and P version, so any doc / suggestion on
> how to configure A/A cinder volume on S version?
>
>
>
>
> Ji Chen
> z Infrastructure as a Service architect
> Phone: 10-82451493
> E-mail: jichenjc at cn.ibm.com
>
>




More information about the openstack-discuss mailing list