Re: Forget close an open relation in ReorderBufferProcessTXN()

On Mon, May 24, 2021 at 12:16 PM osumi.takamichi@fujitsu.com
<osumi.takamichi@fujitsu.com> wrote:
> On Saturday, May 22, 2021 11:58 AM Amit Langote <amitlangote09@gmail.com> wrote:
> > On Sat, May 22, 2021 at 11:00 AM osumi.takamichi@fujitsu.com
> > <osumi.takamichi@fujitsu.com> wrote:
> > > I've checked the core file of v3's failure core and printed the entry
> > > to get more confidence. Sorry for inappropriate measure to verify the
> > solution.
> > >
> > > $1 = {relid = 16388, schema_sent = false, streamed_txns = 0x0,
> > replicate_valid = false, pubactions = {pubinsert = false, pubupdate = false,
> > pubdelete = false, pubtruncate = false}, publish_as_relid = 16388,
> > >   map = 0x7f7f7f7f7f7f7f7f}
> > >
> > > Yes, the process tried to free garbage.
> > > Now, we are convinced that we have addressed the problem. That's it !
> >
> > Thanks for confirming that.
> Langote-san, I need to report another issue.

Thanks for continued testing.

> When I execute make check-world with v6 additionally,
> I've gotten another failure. I get this about once in
> 20 times of make check-world with v6.
>
> The test ended with stderr outputs below.
>
> NOTICE:  database "regression" does not exist, skipping
> make[2]: *** [check] Error 1
> make[1]: *** [check-isolation-recurse] Error 2
> make[1]: *** Waiting for unfinished jobs....
> make: *** [check-world-src/test-recurse] Error 2
> make: *** Waiting for unfinished jobs....
>
> And, I had ./src/test/isolation/output_iso/regression.diffs and regression.out,
> which told me below.
>
> test detach-partition-concurrently-1 ... ok          705 ms
> test detach-partition-concurrently-2 ... ok          260 ms
> test detach-partition-concurrently-3 ... FAILED      618 ms
> test detach-partition-concurrently-4 ... ok         1384 ms
>
> The diffs file showed me below.
>
> diff -U3 /home/k5user/new_disk/repro_fail_v6/src/test/isolation/expected/detach-partition-concurrently-3.out /home/k5user/new_disk/repro_fail_v6/src/test/isolation/output_iso/results/detach-partition-concurrently-3.out
> --- /home/k5user/new_disk/repro_fail_v6/src/test/isolation/expected/detach-partition-concurrently-3.out 2021-05-24 01:22:22.381488295 +0000
> +++ /home/k5user/new_disk/repro_fail_v6/src/test/isolation/output_iso/results/detach-partition-concurrently-3.out       2021-05-24 02:47:08.292488295 +0000
> @@ -190,7 +190,7 @@
>
>  t
>  step s2detach: <... completed>
> -error in steps s1cancel s2detach: ERROR:  canceling statement due to user request
> +ERROR:  canceling statement due to user request
>  step s2detach2: ALTER TABLE d3_listp DETACH PARTITION d3_listp2 CONCURRENTLY;
>  ERROR:  partition "d3_listp1" already pending detach in partitioned table "public.d3_listp"
>  step s1c: COMMIT;
>
> I'm not sure if this is related to the patch or we already have this from OSS HEAD yet.

Hmm, I doubt it would be this patch's fault.  Maybe we still have some
unresolved issues with DETACH PARTITION CONCURRENTLY.  I suggest you
report this in a new thread preferably after you figure that it's
reproducible.

-- 
Amit Langote
EDB: http://www.enterprisedb.com