tarantool / tarantool / 8022
87%
master: 87%

LAST BUILD BRANCH: gh-10776-config-replication-disabled-part-1
DEFAULT BRANCH: master
Ran
Jobs 1
Files 223
Run time 18s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
8022

push

travis-ci

locker 
Rework space truncation

Space truncation that we have now is not atomic: we recreate all indexes
of the truncated space one by one. This can result in nasty failures if
a tuple insertion races with the space truncation and sees some indexes
truncated and others not.

This patch redesigns space truncation as follows:

 - Truncate is now triggered by a new space "option", truncate_count.
   To truncate a space, you need to bump it. As before, space truncation
   is implemented by recreating all of its indexes, but now this is done
   internally in one go, inside the space alter trigger. This makes the
   operation atomic.

 - Although this makes truncation a part of space alter, we don't reuse
   the alter infrastructure, because truncation is quite different: the
   new space must be instantiated before WAL write. We can't instantiate
   it after WAL write, as we do in case of space alter, because this
   would make truncation non-linearizable, breaking replication. To
   understand this, imagine UPSERT({1, 1}, {{'=', 2, 10}}) coming right
   after truncation of a space that contains {1}. One would expect it to
   result in {1, 1}, but if we don't replace the old space with a new
   empty one before WAL write, it might happen that the UPSERT is
   executed before truncate record is written to WAL, resulting in
   {1, 10} present in the space, which is incorrect.

 - New indexes are created with Handler::createIndex method, old indexes
   are destroyed by Handler::~Handler. Neither Handler::commitIndex nor
   Handler::dropIndex are called in case of truncation, in contrast to
   space alter. Since memtx needs to release tuples referenced by old
   indexes, and vinyl needs to log space truncation in the metadata log,
   a new Handler method is introduced, commitTruncateSpace, which is
   passed the old and new spaces. It is called after truncate record has
   successfully been written to WAL.

 - Since Handler::commitTruncateSpace must not fail while vylog write
   obviously may, we need to reuse the technique used by commitIndex and
   dropIndex methods of VinylSpace, namely leave the record we failed to
   write in vylog buffer to be either flushed along with the next write
   or replayed on WAL recovery. To be able to detect if truncation was
   logged while recovering WAL, we introduce a new vylog record type,
   VY_LOG_TRUNCATE_INDEX which takes truncate_count as a key: if on WAL
   recovery index truncate_count happens to be > space truncate_count,
   then it it means that truncation was not logged and we need to log it
   again.

Closes #618
Closes #2060

28539 of 32762 relevant lines covered (87.11%)

1140079.27 hits per line

Jobs
ID Job ID Ran Files Coverage
4 8022.4 (TARGET=coverage) 223
87.11
 Travis Job 8022.4
Source Files on build 8022
Coverage ∆ File Lines Relevant Covered Missed Hits/Line