Upgrading from 7.2.5 to 7.3.3 to mitigate the Datetime.xml problem before the new year.
I have a multi-site indexer cluster, five peers in site1, five peers in site2.
I have a search head cluster, 6 members in site1 and 4 members in site2.
If I can use the site-by-site upgrade option, then I can keep ingesting data and maintain integrity, I never have to bring down all indexers at once. However, this option doesn't seem to account for a search head cluster, where there is also a deployer to worry about.
It seems the other option is to upgrade in tiers. This option accounts for the deployer and I can do a rolling restart of the search head members, but the indexers must be brought down all at once.
Am I missing something in the docs? Or is it acceptable to somehow combine the two by nesting the site-by-site indexer upgrade within the tiered upgrade? Like this:
1. Upgrade the Cluster Master
2. Perform a rolling upgrade of the search head cluster
a. Upgrade a non-captain member
b. Upgrade the other members
c. Upgrade the deployer
d. Finalize the rolling upgrade
3. Upgrade site1 indexers
4. Upgrade site2 indexers
Thanks for any help!
I would use the approach you suggest.
I think I found it. There is a link I missed in the docs for "Perform a rolling upgrade of an indexer cluster".
This document, combined with the links in the OP will work for me I think.
I might be able to get away with bringing down the indexers one site at a time, but not sure. Instead i'll probably go one-by-one before finalizing.