On 01/03/2017 03:44 PM, sebb wrote: > On 3 January 2017 at 12:33, Daniel Gruno <[email protected]> wrote: >> On 01/03/2017 02:12 AM, sebb wrote: >>> The test VM is now running ES 5.x with a basic database. >>> The Apache config appears OK, Javascript is able to invoke lua scripts OK. >>> >>> However a couple of the queries fail: >>> >>> api/pminfo.lua:169: Backend Database returned code 400! >>> >>> {"error":{"root_cause":[{"type":"illegal_argument_exception","reason":"No >>> search type for >>> [scan]"}],"type":"illegal_argument_exception","reason":"No search type >>> for [scan]"},"status":400} >> >> search_type changed in 5.x to mean something else, and should be omitted >> in scan function in elastic.lua, AIUI. >> >>> >>> api/stats.lua:410: Backend Database returned code 400! >>> >>> {"error":{"root_cause":[{"type":"parsing_exception","reason":"no >>> [query] registered for >>> [limit]","line":1,"col":405}],"type":"parsing_exception","reason":"no >>> [query] registered for [limit]","line":1,"col":405},"status":400} >> >> limit is a thing of the past, and the size parameter has changed as well >> (it not denotes the total docs in the next scroll, not per shard/node). >> It might be easier to adjust this for ES 2.3/5.1 and not support earlier >> versions of ElasticSearch going forward. > > +1 > >> I seem to also recall some harmonization needed between hard configured >> limits to results and how the scroll API works. I'll look into that ASAP. > > So what does lists.apache.org use for the queries? > Or does it not use ES 5.x?
There are minor proprietary edits, but most importantly, it doesn't use the word cloud feature, so the issue is not present there. > >>> >>> I have tried the stats.lua query directly in curl and it fails with >>> the same error, so this is not a Lua issue. >>> >>> It looks like the syntax for these particular queries has changed in 5.x. >>> >>> However I was told that lists.a.o was running on 5.x. >>> If so, then I assume it must have had some fixes to the syntax which >>> have not been back-ported to the mainline code? >>> >>> Anyone got any idea what needs fixing? >>> >>
