I have a huge table with a primary key and a btree_gist index. After some years of operation, the performance of selections went constantly down. I assume the gist index is fragmented.

I have a database of sufficient size that it does not fit entirely in RAM, including indexes that also exceed RAM capacity. When performing queries, I observe significant differences in processing time depending on whether the index needs to be read from disk or is already loaded in RAM. I have confirmed using EXPLAIN ANALYZE that the issue stems from index scans.

I have a database of sufficient size that it does not fit entirely in RAM, including indexes that also exceed RAM capacity. When performing queries, I observe significant differences in processing time depending on whether the index needs to be read from disk or is already loaded in RAM. I have confirmed using EXPLAIN ANALYZE that the issue stems from index scans.