[tarantool-patches] [PATCH v4 0/2] force gc on running out of disk space
k.belyavskiy at tarantool.org
Tue Jul 3 16:03:45 MSK 2018
Garbage collector do not delete xlog unless replica do not notify
master with newer vclock. This can lead to running out of disk
space error and this is not right behaviour since it will stop the
Fix it by forcing gc to clean xlogs for replica with highest lag.
Add an error injection and a test.
Split between two different commits to improve readability.
First is a proposal to separate tx_prio thread from tx since they
behave differently (the first does not support yield call).
Second is actually a proposal to use messages from wall thread to
tx thread to notify gc to do not keep logs for the oldest replica.
Changes in V2:
- Promoting error from wal_thread to tx via cpipe.
Changes in V3:
- Delete consumers and only for replicas (but not backup).
Changes in V4:
- Bug fix and small changes according to review.
Konstantin Belyavskiy (2):
replication: rename thread from tx to tx_prio
replication: force gc to clean xdir on ENOSPC err
src/box/box.cc | 1 +
src/box/gc.c | 49 +++++++++
src/box/gc.h | 16 +++
src/box/relay.cc | 1 +
src/box/wal.cc | 44 ++++++--
src/errinj.h | 1 +
src/fio.c | 7 ++
test/box/errinj.result | 2 +
test/replication/kick_dead_replica_on_enspc.result | 121 +++++++++++++++++++++
.../kick_dead_replica_on_enspc.test.lua | 56 ++++++++++
test/replication/suite.ini | 2 +-
11 files changed, 291 insertions(+), 9 deletions(-)
create mode 100644 test/replication/kick_dead_replica_on_enspc.result
create mode 100644 test/replication/kick_dead_replica_on_enspc.test.lua
2.14.3 (Apple Git-98)
More information about the Tarantool-patches