Concurrency Bug in Alexol

[Losk-x]

Reproduce the bug:

deadlock when running:

./build/microbench \ --keys_file=./datasets/covid \ --keys_file_type=binary \ --read=0.7 --insert=0.2 \ --scan=0.1 \ --operations_num=200000000 \ --table_size=-1 \ --init_table_ratio=0.5 \ --thread_num=32 \ --index=alexol

The cause of the bug (I guess):

thread 1 insert and trigger resize/split in node A. (it will be deleted then, so the lock of node A will never release)
thread 2 scan node A (waiting for the lock releasing and keep retry)

A Solution?

additional "deleted" flag to indicate the node is deleted, the scan should retravel from the root.

[00yyyyyyyd]

I encountered a similar issue with a deadlock when running concurrent operations on the alexol index. Have you found a solution to this problem?

[jscblack]

Similar problem encountered, but just in single thread, there cloud also happen. Just bulkload nothing, insert 1,2 and you will stuck at 3.