This may be due to Elasticsearch having run out of disk space for the indexes at one point, and set them to read-only.
I'm not an Elasticsearch expert, but I did run into this before with some other software that uses it, so this was familiar ground once I realized I was having the same issue with CirrusSearch.
The solution in both cases was to free disk space and/or alter the "high water" or "flood-stage" settings in Elasticsearch, then unblock the indexes as prescribed in this Support desk topic. This seems like something that CirrusSearch could detect, and give better error messages about, or even something that Elasticsearch could give better error messages about, but alas. Here we are, googling the error message.
On the topic of the deprecation warnings in elasticsearch_deprecation.log, I wouldn't worry too much about those. That is a warning to the developers of the CirrusSearch extension (that they need to update their code), not you.
Two years too late, but I hope that helps someone. --Ernstkm (talk) 13:14, 4 June 2024 (UTC)