[stein][manila] share migration misconfiguration ?

Douglas viroel at gmail.com
Fri Feb 5 12:30:05 UTC 2021


Hi,

Yes, looks correct to me.
Since you are working on DHSS=False mode, the SVMs should be peered in
advance by the storage administrator. In DHSS=True mode, starting at Train,
this is also managed by Manila.
Only the SnapMirror relationship is at volume level, which is fully handled
by the NetApp driver, so you don't need to worry about it.
By creating a replica of your share, the driver will create the snapmirror
relationship and initialize it [1]. Your replica starts in 'out-of-sync'
replica-state, meaning that the snapmirror is still 'transferring' data.
Manila will periodically check if the replica is already 'in-sync' and will
update the 'replica-state' as soon it gets synchronized with the source
(snapmirrored). You can also request a 'resync' operation at any moment,
through the 'share-replica-resync' operation[2].

Regards,

[1]
https://opendev.org/openstack/manila/src/branch/master/manila/share/drivers/netapp/dataontap/cluster_mode/data_motion.py#L177-L189
[2]
https://docs.openstack.org/api-ref/shared-file-system/#resync-share-replica

On Fri, Feb 5, 2021 at 9:00 AM Ignazio Cassano <ignaziocassano at gmail.com>
wrote:

> Hi Douglas, you are really kind.
> Let my to to recap and please correct if I am wrong:
>
> - manila share on netapp are under svm
> - storage administrator createx a peering between svm source and svm
> destination (or on single share volume ?)
> - I create a manila share with specs replication type  (the share belongs
> to source svm) . In manila.conf source and destination must have the same
> replication domain
> - Creating the replication type it initializes the snapmirror
>
> Is it correct ?
> Ignazio
>
> Il giorno ven 5 feb 2021 alle ore 12:34 Douglas <viroel at gmail.com> ha
> scritto:
>
>> Hi Ignazio,
>>
>> In order to use share replication between NetApp backends, you'll need
>> that Clusters and SVMs be peered in advance, which can be done by the
>> storage administrators once. You don't need to handle any SnapMirror
>> operation in the storage since it is fully handled by Manila and the NetApp
>> driver. You can find all operations needed here [1][2]. If you have CIFS
>> shares that need to be replicated and promoted, you will hit a bug that is
>> being backported [3] at the moment. NFS shares should work fine.
>>
>> If you want, we can assist you on creating replicas for your shares in
>> #openstack-manila channel. Just reach us there.
>>
>> [1]
>> https://docs.openstack.org/manila/latest/admin/shared-file-systems-share-replication.html
>> [2]
>> https://netapp-openstack-dev.github.io/openstack-docs/victoria/manila/examples/openstack_command_line/section_manila-cli.html#creating-manila-share-replicas
>> [3] https://bugs.launchpad.net/manila/+bug/1896949
>>
>> On Fri, Feb 5, 2021 at 8:16 AM Ignazio Cassano <ignaziocassano at gmail.com>
>> wrote:
>>
>>> Hello, thanks for your help.
>>> I am waiting my storage administrators have a window to help me because
>>> they must setup the snapmirror.
>>> Meanwhile I am trying the host assisted migration but it does not work.
>>> The share remains in migrating for ever.
>>> I am sure the replication-dr works because I tested it one year ago.
>>> I had an openstack on site A with a netapp storage
>>> I had another openstack on Site B with another  netapp storage.
>>> The two openstack installation did not share anything.
>>> So I made a replication between two volumes (shares).
>>> I demoted the source share taking note about its export location list
>>> I managed the destination on openstack and it worked.
>>>
>>> The process for replication is not fully handled by openstack api, so  I
>>> should call netapp api for creating snapmirror relationship or ansible
>>> modules or ask help to my storage administrators , right ?
>>> Instead, using share migration, I could use only openstack api: I
>>> understood that driver assisted cannot work in this case, but host assisted
>>> should work.
>>>
>>> Best Regards
>>> Ignazio
>>>
>>>
>>>
>>> Il giorno gio 4 feb 2021 alle ore 21:39 Douglas <viroel at gmail.com> ha
>>> scritto:
>>>
>>>> Hi Rodrigo,
>>>>
>>>> Thanks for your help on this. We were helping Ignazio in
>>>> #openstack-manila channel. He wants to migrate a share across ONTAP
>>>> clusters, which isn't supported in the current implementation of the
>>>> driver-assisted-migration with NetApp driver. So, instead of using
>>>> migration methods, we suggested using share-replication to create a copy in
>>>> the destination, which will use the storage technologies to copy the data
>>>> faster. Ignazio didn't try that out yet, since it was late in his timezone.
>>>> We should continue tomorrow or in the next few days.
>>>>
>>>> Best regards,
>>>>
>>>> On Thu, Feb 4, 2021 at 5:14 PM Rodrigo Barbieri <
>>>> rodrigo.barbieri2010 at gmail.com> wrote:
>>>>
>>>>> Hello Ignazio,
>>>>>
>>>>> If you are attempting to migrate between 2 NetApp backends, then you
>>>>> shouldn't need to worry about correctly setting the data_node_access_ip.
>>>>> Your ideal migration scenario is a driver-assisted-migration, since it is
>>>>> between 2 NetApp backends. If that fails due to misconfiguration, it will
>>>>> fallback to a host-assisted migration, which will use the
>>>>> data_node_access_ip and the host will attempt to mount both shares. This is
>>>>> not what you want for this scenario, as this is useful for different
>>>>> backends, not your case.
>>>>>
>>>>> if you specify "manila migration-start --preserve-metadata True" it
>>>>> will prevent the fallback to host-assisted, so it is easier for you to
>>>>> narrow down the issue with the host-assisted migration out of the way.
>>>>>
>>>>> I used to be familiar with the NetApp driver set up to review your
>>>>> case, however that was a long time ago. I believe the current NetApp driver
>>>>> maintainers will be able to more accurately review your case and spot the
>>>>> problem.
>>>>>
>>>>> If you could share some info about your scenario such as:
>>>>>
>>>>> 1) the 2 backends config groups in manila.conf (sanitized, without
>>>>> passwords)
>>>>> 2) a "manila show" of the share you are trying to migrate (sanitized
>>>>> if needed)
>>>>> 3) the "manila migration-start" command you are using and its
>>>>> parameters.
>>>>>
>>>>> Regards,
>>>>>
>>>>>
>>>>> On Thu, Feb 4, 2021 at 2:06 PM Ignazio Cassano <
>>>>> ignaziocassano at gmail.com> wrote:
>>>>>
>>>>>> Hello All,
>>>>>> I am trying to migrate a share between a netapp backend to another.
>>>>>> Both backends are configured in my manila.conf.
>>>>>> I am able to create share on both, but I am not able to migrate share
>>>>>> between them.
>>>>>> I am using DSSH=False.
>>>>>> I did not understand how host and driver assisted migration work and
>>>>>> what  "data_node_access_ip" means.
>>>>>> The share I want to migrate is on a network (10.102.186.0/24) that I
>>>>>> can reach by my management controllers network (10.102.184.0/24). I
>>>>>> Can mount share from my controllers and I can mount also the netapp SVM
>>>>>> where the share is located.
>>>>>> So in the data_node_access_ip I wrote the list of my controllers
>>>>>> management ips.
>>>>>> During the migrate phase I checked if my controller where manila is
>>>>>> running mounts the share or the netapp SVM but It does not happen.
>>>>>> Please, what is my mistake ?
>>>>>> Thanks
>>>>>> Ignazio
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>> Rodrigo Barbieri
>>>>> MSc Computer Scientist
>>>>> OpenStack Manila Core Contributor
>>>>> Federal University of São Carlos
>>>>>
>>>>>
>>>>
>>>> --
>>>> Douglas Salles Viroel
>>>>
>>>
>>
>> --
>> Douglas Salles Viroel
>>
>

-- 
Douglas Salles Viroel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20210205/8940c70d/attachment-0001.html>


More information about the openstack-discuss mailing list