rockeet
rockeet
## Step 1 start a clean myrocks server ## Step 2 run sysbench on a client ```bash sysbench oltp_write_only prepare --threads=100 --tables=100 --table_size=100000 \ --db-driver=mysql --mysql-host= --mysql-port=3306 \ --mysql-user=test --mysql-password=...
@lth Yes, this is the issue you had documented. This issue is serious as it would yield wrong query plans, could you provide a fix?
> whats the motivation of this change? if dont `ignore_range_deletions`, DB reads will try to construct RangeDelAggregator which waste amount of CPU time without any useful work when `rocksdb_enable_delete_range_for_drop_index` was...
> Thanks for reporting. if you are using new rocksdb version(>=8.8.0), then you can try to disable iterator bound check(rocksdb_check_iterate_bounds = 0); Very thanks!
> I read the diff and I think it is better to submit as a RocksDB PR (https://github.com/facebook/rocksdb) as a read option or a wrapper of Iterator, rather than MyRocks...
Now ToplingDB will disable Intra L0 compaction if `max_level1_subcompactions > 1`, see commit a6a79feaac4d9d0998ec5bc1d68b8f194f6c0f20 > ToplingDB had added `TableFactory::AllowIntraL0Compaction()` for this purpose, then we realized using `max_level1_subcompactions > 1` is...
> I mostly understand this change (I think) but either I am missing something or the title does not agree with the code. Are you syncing the TableProperties to the...
Is there any progress?
> Please look into the test failures - thanks! Fixed, the cause was missing `num_range_deletions`.
Unfortunately, `BeginTransaction()` does not alloc a seq num, this makes WBWIIterator can not take a seq num as `virtual` seq num to be used by MergingIterator's heap comparator.