[heketi-devel] Heketi HA using Kubernetes

[Jose A. Rivera]

In regular usage the heketi service is only used for the creation and
deletion of GlusterFS volumes. That is, only as PV/PVC pairs are
created or deleted. heketi's presence is not needed for when
applications want to mount and read/write to the volumes. If you have
a constant stream of volume create/delete requests then yes you'll
experience a few seconds of service interruption, but the actual I/O
work would not be affected.

--Jose

On Thu, Sep 28, 2017 at 9:28 AM, Gaurav Chhabra
<varuag.chhabra at gmail.com> wrote:
> Hi Talur,
>
>
> Since there will be many applications which will be using Gluster volumes, i
> would really prefer if i could ensure maximum availability. From what i
> understand, since Heketi is managing Gluster and while creating storage
> class, we specify the Heketi REST URL, so if it goes down, i hope
> application pods will be affected, isn't it? If replicas is not feasible,
> what are the options i can explore?
>
>
> Regards,
> Gaurav
>
>
> On Thu, Sep 28, 2017 at 5:39 PM, Raghavendra Talur <rtalur at redhat.com>
> wrote:
>>
>> On Wed, Sep 27, 2017 at 11:15 PM, Gaurav Chhabra
>> <varuag.chhabra at gmail.com> wrote:
>> > Hello Everyone,
>> >
>> >
>> > I want to install Heketi server and client on Kubernetes. To ensure HA,
>> > i
>> > would like to have replica set to 2 for it. However, i am not sure about
>> > the
>> > Heketi db. Does it contain only static data about volumes, nodes etc
>> > (and
>> > changes only when we perform some explicit operations such as add
>> > volumes
>> > etc) OR will it create issues (concurrent write) if i have two replicas
>> > running?
>> >
>>
>> Yes, heketi updates the db only when a operation is performed. But
>> replica set 2 won't work because bolt db opens the db file with read
>> write lock and the second instance will fail to get one. You will end
>> up with 2nd instance in read only mode.
>>
>> If I get it right, you are saying you can't afford to have heketi pod
>> down even for the time duration when kubernetes respins the pod when
>> it detects it is down?
>>
>> >
>> > Thanks,
>> > Gaurav
>> >
>> > _______________________________________________
>> > heketi-devel mailing list
>> > heketi-devel at gluster.org
>> > http://lists.gluster.org/mailman/listinfo/heketi-devel
>> >
>
>
>
> _______________________________________________
> heketi-devel mailing list
> heketi-devel at gluster.org
> http://lists.gluster.org/mailman/listinfo/heketi-devel
>