Kyma High availability

[akuller]

Hello everyone,

This is a silly question, but I can't really find the right answer.

Let's assume we are creating a basic Kyma cluster with three nodes. According to the documentation, they are in different regions. How is a region defined? As with CF, in a different data centre?

The more interesting question is, how are the components updated? Is this done per node? So as not to cripple the entire cluster? Is this then based on region, as with CF? Where can I see the time periods for this, is there a note?

There is therefore no need to set up multiple clusters in different sub-accounts.

If this is the case, Kyma would be the more elegant option than a multi-region setup using CF.

Best regards

[Yogananda]

Hi @akuller 

Maybe Kyma Estimator might provide you some hints on managing CU level
https://kyma-project.github.io/price-calculator/ 

[gabbi]

Hi @akuller ,

Please find details below:

Kyma is highly available inside a region. This means that the 3 nodes gets deployed on 3 different availability zones inside a region.  https://help.sap.com/docs/btp/sap-business-technology-platform/availability-zones-in-kyma-environmen... 

Rearding the components, e.g. the Kyma modules are getting updated on a regular basis. They are zero downtime updates.

The major updates (e.g. Kubernetes version upgrade) follow the BTP maintenance window schedule. https://support.sap.com/en/my-support/systems-installations/cac/maintenance-windows.html?anchorId=se...

Ensuring that your applications are also highly available, you can set up multiple replicas across different availability zones. https://help.sap.com/docs/btp/sap-business-technology-platform/resilient-applications-in-kyma-enviro... 

BR

Gaurav

[akuller]

Hi @gabbi ,

> Rearding the components, e.g. the Kyma modules are getting updated on a regular basis. They are zero downtime updates.

What happens if a node goes completely down due to the update/upgrade? Will the other nodes also be updated? Are the update periods at different times for each data centre/AZ?

> Rearding the components, e.g. the Kyma modules are getting updated on a regular basis. They are zero downtime updates.

I guess it's behind SAP Business Technology Platform? I'm only familiar with this note https://me.sap.com/notes/3459911 , but it doesn't say much, especially the maintenance window issue triggers me. If updates take place in all nodes at the same time, you would need another cluster in a different environment to avoid this. Right?

[gabbi]

In most of the cases, the nodes themselves are updated one by one. The maintenance windows schedules as per region are mentioned in the link shared previously. https://support.sap.com/en/my-support/systems-installations/cac/maintenance-windows.html?anchorId=se...

[akuller]

Hi @gabbi ,

> In most of the cases, the nodes themselves are updated one by one.

I am bothered by the definition, can you clarify in which cases this can not occur? Is there any documentation on this somewhere?

> The maintenance windows schedules as per region are mentioned in the link shared previously. https://support.sap.com/en/my-support/systems-installations/cac/maintenance-windows.html?anchorId=se...

Sorry, I can't find it. I can only find an entry from 2023 in the PDF above. Nothing else. Can you tell me where I need to look?

Regarding SLA, you specify 99.95% for Kyma ( https://assets.cdn.sap.com/agreements/product-use-and-support-terms/service-description-guides/sap-b... ). How can 99.99% be achieved? I cannot find any further information on this. For example, you have a own supplementary clause for SAP Hana Cloud ( https://assets.cdn.sap.com/agreements/product-use-and-support-terms/cls/en/sap-hana-cloud-supplement... ).

[gabbi]

All upgrade scenarios can not be envisioned upfront. Please note that any downtimes are part of the maintenance schedule. To access go to the link https://support.sap.com/en/my-support/systems-installations/cac/maintenance-windows.html?anchorId=se... and then in the search box enter "SAP Business Technology Platform". You might need to scroll to right to see all regions. At present, as you also pointed out the SLA is 99.95%. There are no current plans to provide 99.99%. It may change in the future. For rest, please ensure that your applications are also highly available, you can set up multiple replicas across different availability zones. https://help.sap.com/docs/btp/sap-business-technology-platform/resilient-applications-in-kyma-enviro...

[akuller]

Hi @gabbi ,

The high availability of the own application is not the problem; the problem is rather the availability on the platform. If the customer demands very high availability, how can this be implemented in Kyma if only 99.95% is guaranteed and all nodes may crash during an update?

Thanks for pointing out that it falls under BTP; I thought there was a separate item for it.