diff options
author | Michael Cahill <michael.cahill@mongodb.com> | 2016-07-27 16:18:00 +1000 |
---|---|---|
committer | Alex Gorrod <alexander.gorrod@mongodb.com> | 2016-07-27 16:18:00 +1000 |
commit | 7e1a47dd45735a8ee98aaf24f60406a0bb682359 (patch) | |
tree | b9f5502ee918c8222b1edc9ee31d953f464c4af0 /bench/wtperf/runners/checkpoint_schema_race.wtperf | |
parent | e892427f099623146e0a5343d95ceb6dbb082d0f (diff) | |
download | mongo-7e1a47dd45735a8ee98aaf24f60406a0bb682359.tar.gz |
WT-2798 Fix data consistency bug with table creates during a checkpoint. (#2910)
When logging is disabled, a create operation (and potentially other
metadata updates) could write partially completed checkpoint metadata,
leaving on-disk files inconsistent until the checkpoint completes.
Diffstat (limited to 'bench/wtperf/runners/checkpoint_schema_race.wtperf')
-rw-r--r-- | bench/wtperf/runners/checkpoint_schema_race.wtperf | 20 |
1 files changed, 20 insertions, 0 deletions
diff --git a/bench/wtperf/runners/checkpoint_schema_race.wtperf b/bench/wtperf/runners/checkpoint_schema_race.wtperf new file mode 100644 index 00000000000..ade8e88ee9b --- /dev/null +++ b/bench/wtperf/runners/checkpoint_schema_race.wtperf @@ -0,0 +1,20 @@ +# Check create and drop behavior concurrent with checkpoints (WT-2798). +# Setup a multiple tables and a cache size large enough that checkpoints can +# take a long time. +conn_config="cache_size=8GB,log=(enabled=false),checkpoint=(wait=30)" +table_config="leaf_page_max=4k,internal_page_max=16k,type=file" +icount=10000000 +table_count=100 +table_count_idle=100 +# Turn on create/drop of idle tables, but don't worry if individual operations +# take a long time. +idle_table_cycle=120 +populate_threads=5 +checkpoint_threads=0 +report_interval=5 +# 100 million +random_range=10000000 +run_time=300 +# Setup a workload that dirties a lot of the cache +threads=((count=2,reads=1),(count=2,inserts=1),(count=2,updates=1)) +value_sz=500 |