On 02/27/2018 08:27 AM, Daniel Gruno wrote: > Hi folks, > I am working on removing the reliance on document types inside the > storage, with the aim of making it backwards compatible for older DBs. > > What this means, is that instead of one index with multiple types (we > all knew that wasn't going to last :( ), there will be an index for each > type of document we deal with. New databases will be created using > multiple indices, but older DBs with one index and multiple types will > still be supported by a preflight check. > > I expect this rewrite to take a few days, and I don't expect it to break > much. The DB version will be incremented to 2 on new installations to > mark this shift. Old v1 installations will, for the time being, fall > back to the old way of doing things. > > With regards, > Daniel. >
The base work has now been completed, and new setups will move to a per-type index model. Older 5.x ES instances set up with the earlier setup script will continue to work, but I'll work on an upgrade script for when it becomes vital to change the DB model (while still retaining the previously gathered data). This also means that as of now, the minimum ElasticSearch version has been bumped to 6.x for new installations. There are a few straggling .keyword entries left, that I need to weed out, which may or may not break things on older 5.x DBs, depending on aggregation types. I plan to mitigate this by providing aforementioned upgraded script for 6.x databases.