From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Wed, 27 Mar 2019 13:03:51 +0300 From: Vladimir Davydov Subject: Re: [tarantool-patches] [PATCH v3 2/5] Synchronize lua schema update with space cache Message-ID: <20190327100351.gpo2befo6mulqw2k@esperanza> References: <2c27dc3efab7420737653d3362570136d885dbb4.1553255718.git.georgy@tarantool.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <2c27dc3efab7420737653d3362570136d885dbb4.1553255718.git.georgy@tarantool.org> To: Georgy Kirichenko Cc: tarantool-patches@freelists.org List-ID: On Fri, Mar 22, 2019 at 03:06:07PM +0300, Georgy Kirichenko wrote: > Update lua schema as soon as space cache replace was done instead of > doing this while on_commit trigger executes. In opposite then case > schema changes would not be visible until commit was finished. > > Needed for: #2798 > --- > src/box/alter.cc | 13 ++++++++----- > 1 file changed, 8 insertions(+), 5 deletions(-) > > diff --git a/src/box/alter.cc b/src/box/alter.cc > index daaa9cd57..275e39dd5 100644 > --- a/src/box/alter.cc > +++ b/src/box/alter.cc > @@ -750,8 +750,6 @@ alter_space_commit(struct trigger *trigger, void *event) > op->commit(alter, txn->signature); > } > > - trigger_run_xc(&on_alter_space, alter->new_space); > - > alter->new_space = NULL; /* for alter_space_delete(). */ > /* > * Delete the old version of the space, we are not > @@ -787,6 +785,8 @@ alter_space_rollback(struct trigger *trigger, void * /* event */) > space_swap_triggers(alter->new_space, alter->old_space); > space_swap_fk_constraints(alter->new_space, alter->old_space); > space_cache_replace(alter->new_space, alter->old_space); > + trigger_run(&on_alter_space, alter->old_space); > + > alter_space_delete(alter); > } > > @@ -888,6 +888,7 @@ alter_space_do(struct txn *txn, struct alter_space *alter) > * cache with it. > */ > space_cache_replace(alter->old_space, alter->new_space); > + trigger_run_xc(&on_alter_space, alter->new_space); The comment right above says that this function must not throw an exception here so I guess we should use _xc free variant. Not that it really matters, as on_alter_space trigger can only fail on OOM, which is hardly ever possible. > > /* > * Install transaction commit/rollback triggers to either > @@ -1388,7 +1389,6 @@ on_drop_space_commit(struct trigger *trigger, void *event) > { > (void) event; > struct space *space = (struct space *)trigger->data; > - trigger_run_xc(&on_alter_space, space); > space_delete(space); > } > > @@ -1403,6 +1403,7 @@ on_drop_space_rollback(struct trigger *trigger, void *event) > (void) event; > struct space *space = (struct space *)trigger->data; > space_cache_replace(NULL, space); > + trigger_run(&on_alter_space, space); > } > > /** > @@ -1412,8 +1413,7 @@ static void > on_create_space_commit(struct trigger *trigger, void *event) > { > (void) event; > - struct space *space = (struct space *)trigger->data; > - trigger_run_xc(&on_alter_space, space); > + (void) trigger; > } The function is not used now. I guess it's okay to leave it, just in case we want to extend it in the future. > > /** > @@ -1429,6 +1429,7 @@ on_create_space_rollback(struct trigger *trigger, void *event) > (void) event; > struct space *space = (struct space *)trigger->data; > space_cache_replace(space, NULL); > + trigger_run(&on_alter_space, space); > space_delete(space); > } > > @@ -1672,6 +1673,7 @@ on_replace_dd_space(struct trigger * /* trigger */, void *event) > * execution on a replica. > */ > space_cache_replace(NULL, space); > + trigger_run_xc(&on_alter_space, space); Agagin, we must no throw exceptions here. > /* > * Do not forget to update schema_version right after > * inserting the space to the space_cache, since no > @@ -1764,6 +1766,7 @@ on_replace_dd_space(struct trigger * /* trigger */, void *event) > * execution on a replica. > */ > space_cache_replace(old_space, NULL); > + trigger_run_xc(&on_alter_space, old_space); > /* > * Do not forget to update schema_version right after > * deleting the space from the space_cache, since no I think we should've fixed on_alter_sequence trigger in the scope of this issue, too. The patch has been pushed to master, 2.1, and 1.10 by Kostja. I guess we can address the minor hitches I pointed out above in a separate 'review-fix' patch.