Unable to index large tag values

[jgpruitt]

Seeing a fair number of these in the opentelemetry-demo

UNIQUE CONSTRAINT, btree (key, value) INCLUDE (id, key_id) on tag partitions

STATEMENT:  SELECT ps_trace.put_tag($1, $2, $3::ps_trace.tag_type)
ERROR:  index row size 4608 exceeds btree version 4 maximum 2704 for index "tag_63_key_value_id_key_id_key"
DETAIL:  Index row references tuple (0,37) in relation "tag_63".
HINT:  Values larger than 1/3 of a buffer page cannot be indexed.
	Consider a function index of an MD5 hash of the value, or use full text indexing.
CONTEXT:  SQL statement "INSERT INTO _ps_trace.tag as t (tag_type, key_id, key, value)
	        SELECT
	            _tag_type,
	            k.id,
	            _key,
	            _value
	        FROM _ps_trace.tag_key k
	        WHERE k.key = _key
	        ON CONFLICT (key, value) DO
	        UPDATE SET tag_type = t.tag_type | EXCLUDED.tag_type
	        WHERE t.tag_type & EXCLUDED.tag_type = 0"
	PL/pgSQL function put_tag(tag_k,tag_v,tag_type) line 12 at SQL statement

[cevian]

what is this index used for?

[jgpruitt]

The primary use is to prevent duplicate entries of the same key-value pair.

[jgpruitt]

I guess we COULD use a hash of the value to put a unique index on, but that's annoying. Not sure what else we could do though.

[cevian]

Closing in favor of timescale/promscale_extension#128