Re: BUG #17245: Index corruption involving deduplicated entries

On 30.10.2021 5:37, Andres Freund wrote:
> The problem is that parallel vacuum doesn't have correct handling for small
> indexes, when there are other indexes that are big enough to use parallelism.

Index sizes in use (pg_total_relation_size), btree only, with only the 
largest, the smallest, and those reported by pg_amcheck listed:

┌───────────────────────────────────┬──────────────┬────────────┐
│               index               │ index_method │ index_size │
├───────────────────────────────────┼──────────────┼────────────┤
│  pl_backlinks_namespace           │ btree        │ 18 MB      │
│  pl_namespace                     │ btree        │ 16 MB      │
│  wl_user                          │ btree        │ 6416 kB    │
│  …                                │ …            │ …          │
│  page_redirect_namespace_len      │ btree        │ 544 kB     │
│  …                                │ …            │ …          │
│  page_len                         │ btree        │ 488 kB     │
│  …                                │ …            │ …          │
│  transcode_key_idx                │ btree        │ 192 kB     │
│  …                                │ …            │ …          │
│  page_main_title                  │ btree        │ 144 kB     │
│  …                                │ …            │ …          │
│  ir_type_value                    │ btree        │ 8192 bytes │
│  ipc_hex_time                     │ btree        │ 8192 bytes │
└───────────────────────────────────┴──────────────┴────────────┘


-- 
K. R.