Re: brininsert optimization opportunity

On 12/11/23 16:00, Alexander Lakhin wrote:
> Hello Tomas and Soumyadeep,
> 
> 25.11.2023 23:06, Tomas Vondra wrote:
>> I've done a bit more cleanup on the last version of the patch (renamed
>> the fields to start with bis_ as agreed, rephrased the comments / docs /
>> commit message a bit) and pushed.
> 
> Please look at a query, which produces warnings similar to the ones
> observed upthread:
> CREATE TABLE t(a INT);
> INSERT INTO t SELECT x FROM generate_series(1,10000) x;
> CREATE INDEX idx ON t USING brin (a);
> REINDEX index CONCURRENTLY idx;
> 
> WARNING:  resource was not closed: [1863] (rel=base/16384/16389,
> blockNum=1, flags=0x93800000, refcount=1 1)
> WARNING:  resource was not closed: [1862] (rel=base/16384/16389,
> blockNum=0, flags=0x93800000, refcount=1 1)
> 
> The first bad commit for this anomaly is c1ec02be1.
> 

Thanks for the report. I haven't investigated what the issue is, but it
seems we fail to release the buffers in some situations - I'd bet we
fail to call the cleanup callback in some place, or something like that.
I'll take a look tomorrow.

> May be you would also want to fix in passing some typos/inconsistencies
> introduced with recent brin-related commits:
> s/bs_blkno/bt_blkno/
> s/emptry/empty/
> s/firt/first/
> s/indexinsertcleanup/aminsertcleanup/
> s/ maxRange/ nextRange/
> s/paga /page /
> 

Definitely. Thanks for noticing those!


regards

-- 
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company