bug: abort when query through index on dataset arguana

[kemingy]

server log:

2024-12-26 22:50:30.218 +08 [2059266] error:  column reference "id" is ambiguous at character 8
2024-12-26 22:50:30.218 +08 [2059266] statement:  select id, corpus.embedding <&> queries.embedding from corpus, queries where corpus.id = queries.id;
2024-12-26 22:52:50.777 +08 [2066408] log:  skipping vacuum of "corpus" --- lock not available
trap: failed assert("interruptholdoffcount > 0"), file: "lwlock.c", line: 1849, pid: 2066300
postgres: user vchord_bm25 172.17.0.2(34150) select(exceptionalcondition+0x54) [0x9d86a4]
postgres: user vchord_bm25 172.17.0.2(34150) select(lwlockrelease+0x5e5) [0x88fc15]
postgres: user vchord_bm25 172.17.0.2(34150) select(unlockreleasebuffer+0xa) [0x86594a]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0xa1ed6) [0x7fd36d2a1ed6]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0xa0749) [0x7fd36d2a0749]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0xa6f9c) [0x7fd36d2a6f9c]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0x62950) [0x7fd36d262950]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0x63848) [0x7fd36d263848]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0xa45e7) [0x7fd36d2a45e7]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0x423ac) [0x7fd36d2423ac]
/home/user/.pgrx/17.0/pgrx-install/lib/postgresql/vchord_bm25.so(+0xbeb67) [0x7fd36d2beb67]
postgres: user vchord_bm25 172.17.0.2(34150) select(index_getnext_tid+0x4a) [0x58817a]
postgres: user vchord_bm25 172.17.0.2(34150) select(index_getnext_slot+0x6b) [0x58837b]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x70e72c]
postgres: user vchord_bm25 172.17.0.2(34150) select(execscan+0xe6) [0x6f2726]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x7105f3]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x71a664]
postgres: user vchord_bm25 172.17.0.2(34150) select(standard_executorrun+0x15a) [0x6e8aaa]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x8a4ddf]
postgres: user vchord_bm25 172.17.0.2(34150) select(portalrun+0x240) [0x8a64a0]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x8a24c2]
postgres: user vchord_bm25 172.17.0.2(34150) select(postgresmain+0x1550) [0x8a3e10]
postgres: user vchord_bm25 172.17.0.2(34150) select(backendmain+0x4d) [0x89eb8d]
postgres: user vchord_bm25 172.17.0.2(34150) select(postmaster_child_launch+0xb0) [0x8048f0]
postgres: user vchord_bm25 172.17.0.2(34150) select() [0x8084b0]
postgres: user vchord_bm25 172.17.0.2(34150) select(postmastermain+0xdcb) [0x80a11b]
postgres: user vchord_bm25 172.17.0.2(34150) select(main+0x1d9) [0x51bcb9]
/lib64/libc.so.6(+0x3248) [0x7fd38761a248]
/lib64/libc.so.6(__libc_start_main+0x8b) [0x7fd38761a30b]
postgres: user vchord_bm25 172.17.0.2(34150) select(_start+0x25) [0x51bf95]
2024-12-26 22:52:52.432 +08 [2059253] log:  server process (pid 2066300) was terminated by signal 6: aborted
2024-12-26 22:52:52.432 +08 [2059253] detail:  failed process was running: select q.id as qid, c.id, c.score from queries q, lateral (select id, corpus.embedding <&> q.embedding as score from corpus order by score limit 10) c;
2024-12-26 22:52:52.432 +08 [2059253] log:  terminating any other active server processes
2024-12-26 22:52:52.433 +08 [2059253] log:  all server processes terminated; reinitializing
2024-12-26 22:52:52.454 +08 [2066442] LOG:  database system was interrupted; last known up at 2024-12-26 22:49:58 +08
2024-12-26 22:52:52.947 +08 [2066442] LOG:  database system was not properly shut down; automatic recovery in progress
2024-12-26 22:52:52.953 +08 [2066442] LOG:  redo starts at D/6EB89DE8
2024-12-26 22:52:53.065 +08 [2066442] LOG:  invalid record length at D/70F3C9B8: expected at least 24, got 0
2024-12-26 22:52:53.065 +08 [2066442] LOG:  redo done at D/70F3C990 system usage: CPU: user: 0.03 s, system: 0.07 s, elapsed: 0.11 s
2024-12-26 22:52:53.085 +08 [2066443] LOG:  checkpoint starting: end-of-recovery immediate wait
2024-12-26 22:52:53.787 +08 [2066443] LOG:  checkpoint complete: wrote 16287 buffers (99.4%); 0 WAL file(s) added, 2 removed, 0 recycled; write=0.151 s, sync=0.494 s, total=0.709 s; sync files=69, longest=0.046 s, average=0.008 s; distance=36554 kB, estimate=36554 kB; lsn=D/70F3C9B8, redo lsn=D/70F3C9B8

[silver-ymz]

Postgres can only hold up to 200 locks at the same time. When the length of query vector is more than 100, blockwand will open 200 pages at once. This is an algorithmic problem that needs to be solved.