Jeff Janes <jeff.janes@gmail.com> writes:
> I was looking into a possible scalability problem with GIN indexes under
> concurrent insert, but instead I found an uncharacterized bug. One of the
> processes will occasionally throw an error "ERROR: buffer 10112 is not
> owned by resource owner Portal" where the buffer number changes from run to
> run.
I am able to reproduce this in HEAD (8bfb231b4) on a not-that-big
machine (M2 Mac Mini):
pgbench: error: client 1 script 0 aborted in command 0 query 0: ERROR: buffer pin 0x1cab is not owned by resource
ownerPortal
pgbench: error: Run was aborted; the above results are incomplete.
FTR, I used the attached not-so-ubuntu-specific script, and a
very vanilla assert-enabled build.
However, it seems like there might be more than one bug. My first
attempt died like this after a few minutes:
TRAP: failed Assert("ref != NULL"), File: "bufmgr.c", Line: 2447, PID: 7177
0 postgres 0x0000000102caab74 ExceptionalCondition + 108
1 postgres 0x0000000102b46490 UnpinBuffer + 528
2 postgres 0x000000010282bae4 ginFinishSplit + 1700
3 postgres 0x00000001028368ac ginEntryInsert + 904
4 postgres 0x0000000102837194 gininsert + 472
5 postgres 0x00000001029bf77c ExecInsertIndexTuples + 552
6 postgres 0x00000001029ed0d4 ExecInsert + 1948
7 postgres 0x00000001029ebbfc ExecModifyTable + 1252
8 postgres 0x00000001029c0dc4 standard_ExecutorRun + 268
9 postgres 0x0000000102b86450 ProcessQuery + 160
10 postgres 0x0000000102b85afc PortalRunMulti + 396
11 postgres 0x0000000102b85478 PortalRun + 476
12 postgres 0x0000000102b8453c exec_simple_query + 1312
13 postgres 0x0000000102b82250 PostgresMain + 3452
That was while running a build of commit 9ba9c7074 from 25 October.
After updating to current HEAD (8bfb231b4), I've not yet reproduced it.
But I suspect that's just bad luck, because this doesn't look like
anything we fixed in the last couple weeks.
regards, tom lane
#!/bin/bash
dropdb test
createdb test
export PGDATABASE=test
psql -c '
create table j (id bigserial, j jsonb);
create index on j using gin (j);
alter index j_j_idx set (fastupdate = off);
'
psql -c '
alter system set synchronous_commit to off;
'
pg_ctl reload
for g in `seq 1 100`; do
for f in `seq 1 10`; do
echo "trying with $f clients"
psql -c 'truncate j'
pgbench -n -f <(echo 'insert into j (j) select jsonb_object_agg(x::text, left(md5(random()::text),5)) from
generate_series(1,10)f(x);') -T 60 -c $f -j $f | grep tps
done
done
