Re: Get stuck when dropping a subscription during synchronizing table

On Tue, Jun 20, 2017 at 10:55 AM, Peter Eisentraut
<peter(dot)eisentraut(at)2ndquadrant(dot)com> wrote:
> On 6/1/17 13:37, Petr Jelinek wrote:
>> On 01/06/17 17:32, Masahiko Sawada wrote:
>>> On Thu, May 25, 2017 at 5:29 PM, tushar <tushar(dot)ahuja(at)enterprisedb(dot)com> wrote:
>>>> After applying all your patches, drop subscription no more hangs while
>>>> dropping subscription but there is an error "ERROR: tuple concurrently
>>>> updated" in the log file of
>>>> standby.
>>>
>>> I tried to reproduce this issue with latest four patches I submit but
>>> it didn't happen. I guess this issue doesn't related to the issues
>>> reported on this thread and another factor might be cause of this
>>> issue. It would be good to test the same again with latest four
>>> patches or after solved current some open items.
>>
>> That's because your additional patch kills the workers that do the
>> concurrent update. While that's probably okay, I still plan to look into
>> making the subscription and state locking more robust.
>
> It seems we have gotten off track here a bit. What is the current
> proposal to fix "tuple concurrently updated" during DROP SUBSCRiPTION?

I think there is no proposal for it so far. The current proposal is to
fix this problem during ALTER SUBSCRIPTION.

> It seems to me we could just take a stronger lock around
> RemoveSubscriptionRel(), so that workers can't write in there concurrently.
>

Since we reduced the lock level of updating pg_subscription_rel by
commit 521fd4795e3e the same deadlock issue will appear if we just
take a stronger lock level.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center