[Docs] Provide best practices for the safest way of upgrading Kibana

[rwaight]

Describe the feature:
Last week we discussed our upgrade documentation and ways we can improve the upgrade experience; for larger deployments, the proper order becomes important.

While we provide this high level overview for upgrading Kibana, we should also include best practices around the safest way to upgrade Kibana, especially in larger deployments.

For overall Elastic Stack upgrades, I've opened elastic/stack-docs#537 for best practices to be included.

[elasticmachine]

Pinging @elastic/kibana-docs

[KOTungseth]

@joshdover can you help us with one?

[joshdover]

I'm not sure I have any additional guidance here that isn't already outlined in the Stack Upgrade docs.

I think the most important thing we could emphasize is that Kibana does not currently support "rolling upgrades." This means that before bringing up a Kibana instance of the next version, all older instances need to be shut down first. Another way to put this would be: you can't have multiple Kibana nodes running different versions at the same time against the same .kibana index.

[KOTungseth]

@joshdover in the current Kibana upgrade docs, it says, "The recommended path is to upgrade to 6.7 before upgrading to 7.0. This makes it easier to identify the changes you need to make to upgrade and enables you to perform a rolling upgrade with no downtime."

Since Kibana does not currently support rolling upgrades, should this be removed from the docs?

[joshdover]

@KOTungseth Yes, I think we should add a section about this. In particular, this is what we need to get across:

• Kibana does not support running more than one version of Kibana against the same Elasticsearch index.
• When upgrading, you must shutdown all nodes of the old version before bringing up any nodes of the next version.

"The recommended path is to upgrade to 6.7 before upgrading to 7.0. This makes it easier to identify the changes you need to make to upgrade and enables you to perform a rolling upgrade with no downtime."

This still seems pertinent, but I think we should clarify that we mean this "enables you to perform a rolling upgrade of Elasticsearch with no downtime." (not Kibana).

[rwaight]

Thank you very much @joshdover and @KOTungseth!!

[b-deam]

Hey @joshdover and @KOTungseth 👋!

The important step of ensuring that all Kibana instances are shutdown prior to an upgrade is missing from the documentation for version 6.8.

6.8:

7.x:

Is it possible to backport this PR/add the upgrade advice to the 6.8 "before you begin" page?

I'm happy to raise a seperate PR to the 6.8 branch if that's determined the best way forward.