Open Stack

On 15/05, Souza, Nahim wrote:
> Hi, Raghavendra,
>
> Just sharing my experience, I started working on A/A support for NetApp NFS driver, and I followed the same steps you summarized.
> Besides that, I think the effort is to understand/test if any of the driver features might break in the A/A environment.
>
> If anyone knows about anything else we should test, I would be happy to know too.
>
> Regards,
> Nahim Souza.

Hi,

I see that there is no mention of deploying and configuring the DLM,
which is necessary for the critical sections between different hosts.

For reference, most drivers require code changes. A very common location
were drivers need code changes is when creating the host entities in the
storage array. I see that the 3PAR driver has already changed that lock
to use the DLM locks [1].

If a driver doesn't support replication then there is no need to split
the failover_host method.

Cheers,
Gorka.

[1]: https://github.com/openstack/cinder/blob/7bca35c935cf8566bcf9e4874be78174cb0b6df5/cinder/volume/drivers/hpe/hpe_3par_fc.py#L169

>
>
> From: U T, Raghavendra <raghavendra-uddhav.tilay at hpe.com>
> Sent: Monday, May 8, 2023 09:18
> To: openstack-discuss at lists.openstack.org
> Subject: [cinder][dev] Add support in driver - Active/Active High Availability
>
> NetApp Security WARNING: This is an external email. Do not click links or open attachments unless you recognize the sender and know the content is safe.
>
>
> Hi,
>
> We wish to add Active/Active High Availability to:
> 1] HPE 3par driver - cinder/cinder/volume/drivers/hpe/hpe_3par_common.py
> 2] Nimble driver - cinder/cinder/volume/drivers/hpe/nimble.py
>
> Checked documentation at
> https://docs.openstack.org/cinder/latest/contributor/high_availability.html
> https://docs.openstack.org/cinder/latest/contributor/high_availability.html#cinder-volume
> https://docs.openstack.org/cinder/latest/contributor/high_availability.html#enabling-active-active-on-drivers
>
> Summary of steps:
> 1] In driver code, set SUPPORTS_ACTIVE_ACTIVE = True
> 2] Split the method failover_host() into two methods: failover() and failover_completed()
> 3] In cinder.conf, specify cluster name in [DEFAULT] section
> cluster = <cluster_name>
> 4] Configure atleast two nodes in HA and perform testing
>
> Is this sufficient or anything else required ?
>
> Note: For Nimble driver, replication feature is not yet added.
> So can the above step 2 be skipped?
>
> Appreciate any suggestions / pointers.
>
> Regards,
> Raghavendra Tilay.
>
>
>