PDA

View Full Version : Manually driven Cluster



irekalde_itp
27-Nov-2014, 17:09
Hi,
Maybe is a non-sense, but we had very bad experiences with other clusters software.
And we know that we want to do is essentially against the HA idea. We would like to know if it possible to get fully manually driven HA Cluster to avoid unwanted failover.
Do you know if it is possible to do it?
Best Regard,
Iņigo Rekalde

ab
28-Nov-2014, 04:57
I believe if you set the primitive resource's is-managed attribute to
false that it will not be handled by the cluster in any automatic way:

https://www.suse.com/documentation/sle_ha/book_sleha/data/sec_ha_configuration_basics_resources.html

With that written, why would you want to do this? The more details, the
better we can help you find the best solution possible.

--
Good luck.

If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...

irekalde_itp
28-Nov-2014, 09:43
Thanks for your reply.
We are migrating our SAP Business Suite to HANA. I don’t know if you are familiar with SAP HANA, as a very roughly explanation I can tell you that SAP HANA is an in-memory database and platform.

Now we have SAP Business Suite based on MS Windows and SQL Server in a virtual vmware environment. And we have Disaster Recovery solution based on storage replication and vmware SRM in two datacenters.
We are going to change our virtual machines by two physical machine (one in each datacenter) running SLES for SAP and SAP HANA on them.
We are going to have two node (in separated datacenters) for our productive environment replicated by HANA System replication, one node as the primary and the other one as a standby/backup node.

We want to get a High Availability / Disaster Recovery Solution and we don´t know if we should go to manually driven solution (using scripts) or use the Suse HA Solution with an particular configuration.
In the past we don´t have very good experiences with automatic cluster software. We prefer to have the control when the system should do the takeover to avoid unwanted takeovers.
We know that if we implement a cluster solution it should be to use in an automatic mode. Our experience tell us that we are not going to need switch between nodes more than twice in a year. But in the past when we had introduced clusters it had generated more problems that ones that had solved. With our network and other infrastructure we think that we are more vulnerable that a failure on it provoke an unwanted takeover than the real needs to have a desired takeover. In the past maybe by misconfiguration or by other reasons we have more unwanted takeover than the desired ones.
Our business requirements allow us to have the system down until an operator decided to do a manual takeover.

We would like to know if you think that it is possible to configure the Suse HA Solution with this requirements using “is-managed” or other configuration:
- The Cluster will never take the decision of the takeover.
- A human operator will take the decision of the takeover.
- The operator with the right privileges after analyzing the situation has as simple as possible to do the takeover (ideally run one command or script).

If it possible we prefer to implement the Suse HA Solution than using a completely manual scripting solution. We know that SUSE and SAP are working hard together to get best solution to get a HA Solution for SAP HANA. And in case that in the future we feel more comfortable with this solution (and with our infrastructure) we could reconfigure to be automatically driven instead manually driven.

I hope that I have explained our situation and the requirements.
Any idea or comment would be appreciated.

Best regards,
Iņigo





I believe if you set the primitive resource's is-managed attribute to
false that it will not be handled by the cluster in any automatic way:

https://www.suse.com/documentation/sle_ha/book_sleha/data/sec_ha_configuration_basics_resources.html

With that written, why would you want to do this? The more details, the
better we can help you find the best solution possible.

--
Good luck.

If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...

ab
01-Dec-2014, 04:38
On 11/28/2014 01:44 AM, irekalde itp wrote:
>
> Thanks for your reply.
> We are migrating our SAP Business Suite to HANA. I donīŋŊt know if you are
> familiar with SAP HANA, as a very roughly explanation I can tell you
> that SAP HANA is an in-memory database and platform.

From a high-level yes. Being in-memory, starting up probably takes a huge
chunk of time, especially considering the scales of data upon which SAP
HANA operates, so an automatic, accidental failover is a problem. Why
that would ever happen, though, other than a misconfiguration of some
kind, I cannot fathom.

> We want to get a High Availability / Disaster Recovery Solution and we
> donīŋŊt know if we should go to manually driven solution (using scripts)
> or use the Suse HA Solution with an particular configuration.
> In the past we donīŋŊt have very good experiences with automatic cluster
> software. We prefer to have the control when the system should do the
> takeover to avoid unwanted takeovers.
> We know that if we implement a cluster solution it should be to use in
> an automatic mode. Our experience tell us that we are not going to need
> switch between nodes more than twice in a year. But in the past when we
> had introduced clusters it had generated more problems that ones that
> had solved. With our network and other infrastructure we think that we
> are more vulnerable that a failure on it provoke an unwanted takeover
> than the real needs to have a desired takeover. In the past maybe by
> misconfiguration or by other reasons we have more unwanted takeover than
> the desired ones.
> Our business requirements allow us to have the system down until an
> operator decided to do a manual takeover.

It sounds like implementing the cluster as mentioned previously may be
what you need, then. If your business does not care about a bit of
unexpected downtime while somebody investigates and eventually, manually,
flips the switch to change nodes, then go for that. At some point if you
want to change things, then do so within the cluster configuration.
HANA's stuff, as I understand it, probably does not qualify as being as
mission-critical as some other systems, though I am sure that is an
overly-broad statement and depends a lot on the type of business.

> We would like to know if you think that it is possible to configure the
> Suse HA Solution with this requirements using īŋŊis-managedīŋŊ or other
> configuration:
> - The Cluster will never take the decision of the takeover.
> - A human operator will take the decision of the takeover.
> - The operator with the right privileges after analyzing the situation
> has as simple as possible to do the takeover (ideally run one command or
> script).

Yes, I think so.

> If it possible we prefer to implement the Suse HA Solution than using a
> completely manual scripting solution. We know that SUSE and SAP are
> working hard together to get best solution to get a HA Solution for SAP
> HANA. And in case that in the future we feel more comfortable with this
> solution (and with our infrastructure) we could reconfigure to be
> automatically driven instead manually driven.

Exactly; start out manual, flip to automatic when ready.

--
Good luck.

If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...