[tarantool-patches] Re: [PATCH v2 3/5] Enforce applier out of order protection
Vladimir Davydov
vdavydov.dev at gmail.com
Tue Jan 29 15:00:43 MSK 2019
On Tue, Jan 29, 2019 at 01:30:40PM +0300, Георгий Кириченко wrote:
> On Monday, January 28, 2019 3:09:01 PM MSK Vladimir Davydov wrote:
> > On Tue, Jan 22, 2019 at 01:31:11PM +0300, Georgy Kirichenko wrote:
> > > Do not skip row until the row is not processed by other appliers.
> >
> > Looks like a fix for
> >
> > https://github.com/tarantool/tarantool/issues/3568
> >
> > Worth adding a test?
> >
> > > Prerequisite #980
> > > ---
> > >
> > > src/box/applier.cc | 35 ++++++++++++++++++-----------------
> > > 1 file changed, 18 insertions(+), 17 deletions(-)
> > >
> > > diff --git a/src/box/applier.cc b/src/box/applier.cc
> > > index 87873e970..148c8ce5a 100644
> > > --- a/src/box/applier.cc
> > > +++ b/src/box/applier.cc
> > > @@ -504,6 +504,22 @@ applier_subscribe(struct applier *applier)
> > >
> > > applier->lag = ev_now(loop()) - row.tm;
> > > applier->last_row_time = ev_monotonic_now(loop());
> > >
> > > + struct replica *replica = replica_by_id(row.replica_id);
> > > + struct latch *latch = (replica ? &replica->order_latch :
> > > + &replicaset.applier.order_latch);
> > > + /*
> > > + * In a full mesh topology, the same set
> > > + * of changes may arrive via two
> > > + * concurrently running appliers. Thanks
> > > + * to vclock_follow() above, the first row
> >
> > I don't see any vclock_follow() above. Please fix the comment.
> >
> > > + * in the set will be skipped - but the
> > > + * remaining may execute out of order,
> > > + * when the following xstream_write()
> > > + * yields on WAL. Hence we need a latch to
> > > + * strictly order all changes which belong
> > > + * to the same server id.
> > > + */
> > > + latch_lock(latch);
> > >
> > > if (vclock_get(&replicaset.applier.vclock,
> > >
> > > row.replica_id) < row.lsn) {
> > >
> > > if (row.replica_id == instance_id &&
> >
> > AFAIU this patch makes replicaset.applier.vclock, introduced by the
> > previous patch, useless.
> You are right now, but I plan to release this latch just before commit in case
> of parallel applier.
Then let's please introduce applier.vclock when we get to implement
parallel applier, because currently I can't say for sure whether we
really need it or not.
More information about the Tarantool-patches
mailing list