I have a multisite (specifically, two site) indexer cluster running 7.3, and a search head cluster with search affinity disabled per the recommendation here. I obviously need to upgrade all of those tiers to 8.1 very soon, given the end-of-support dates for 7.3 and 8.0. Ideally I would like to perform that upgrade without an outage.
The best option for this appears to be a site-by-site upgrade, especially as I happen not to have any metrics indexes. Those instructions state that I must first upgrade my cluster manager node to 8.0, then upgrade the indexers in one site to 8.0, and then upgrade the search heads in that same site to 8.0 (then repeat for the other site, then repeat the whole thing to get to 8.1). How does this apply when the search heads are all set to site0?
In contrast, the instructions for upgrading each tier separately say to upgrade the manager, then the search heads, then the indexers.
In what order should I do this upgrade? Do the search heads also have to go from version 7.3 to 8.0 to 8.1 like the cluster manager and indexers do?
Have you engaged Splunk support? This is a fairly complex operation on it's own, but adding in a major version upgrade increases the complexity/risk. If this is a prod environment I would suggest having support help you create a process specific to your environment.