From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Fri, 14 Jun 2019 11:10:08 +0300 From: Vladimir Davydov Subject: Re: [tarantool-patches] [PATCH v3 10/14] core: latch_unlock_external routine Message-ID: <20190614081008.zyuol6q6hn2qxlff@esperanza> References: <98f0889a17245be64f1753012419e92c7ceef4e7.1560112747.git.georgy@tarantool.org> <20190613142735.f4aoatxq7kxnlwps@esperanza> <2220504.ES0OFZsnVI@home.lan> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <2220504.ES0OFZsnVI@home.lan> To: =?utf-8?B?0JPQtdC+0YDQs9C40Lkg0JrQuNGA0LjRh9C10L3QutC+?= Cc: tarantool-patches@freelists.org List-ID: On Thu, Jun 13, 2019 at 10:38:12PM +0300, Георгий Кириченко wrote: > On Thursday, June 13, 2019 5:27:35 PM MSK Vladimir Davydov wrote: > > On Sun, Jun 09, 2019 at 11:44:39PM +0300, Georgy Kirichenko wrote: > > > > Again, looks like a temporary hack to me. I would rather not commit it. > > The right way would be getting rid of the schema_lock. I know that you > > have tried, but failed because of vylog. I'll look what we can do about > > it. > I failed to remove the latch from checkpoint not from ddl. Removing the latch > from alter.cc doesn't look possible in the next 2-3 month. I'll look what we can do about this - AFAICS there are no problems with memtx. We just need to tweak vinyl somehow. Anyway, rather than introducing a new fiber, we can use 'tx' endpoint instead of 'tx_prio' endpoint, just as we used to. It would be equivalent, but look more straightforward AFAIU. That would be a kind of regression though so I would prefer to avoid that if possible and only use it as a last resort.