Baseline Experiments. #345
Conversation
There was a problem hiding this comment.
Thanks @amlatyrngom. Can you test this out on the 100 GB dataset with the scale down workload? experiments/15-e2e-scenarios-v2/scale_down/run_workload.sh. The script assumes it is being run by a tool called Conductor. The easiest way to run it without the tool for testing is to set the COND_OUT environment variable to a path where you want to store the results.
| @pytest.mark.skip( | ||
| reason="TODO(Amadou): This is failing even I haven't changed it. Flaky test?" | ||
| ) |
There was a problem hiding this comment.
This test should be fixed now - you can remove this decorator and the test should pass.
| @@ -0,0 +1,100 @@ | |||
| # See workloads/cross_db_benchmark/benchmark_tools/tidb/README.md | |||
|
|
|||
There was a problem hiding this comment.
Please move this file into workloads/IMDB_extended
| cursor = self._cursor | ||
| # Exec | ||
| cursor.execute(query) | ||
| if cursor.rowcount is None or cursor.rowcount <= 0 or not(query.strip().lower().startswith("SELECT")): |
There was a problem hiding this comment.
query.strip().lower().startswith("SELECT") is always going to be false?
| # Get cursor | ||
| if self._cursor is None: | ||
| had_cursor = False | ||
| cursor = self._conn.cursor() | ||
| else: | ||
| had_cursor = True | ||
| cursor = self._cursor | ||
| # Exec | ||
| cursor.execute(query) |
There was a problem hiding this comment.
Are these changes to support a MySQL connector? If so, please just create another Database class that is specialized for MySQL instead of modifying this class.
| if cursor.rowcount is None or cursor.rowcount <= 0 or not(query.strip().lower().startswith("SELECT")): | ||
| rows = [] | ||
| else: | ||
| print(f"Rows: {cursor.rowcount}. Q: {query}") |
There was a problem hiding this comment.
nit: Please use logger.debug() instead of print(). Otherwise this will get very verbose during a workload run.
| dataset_type = "20gb" | ||
| else: | ||
| dataset_type = "original" | ||
| worker = TransactionWorker(worker_idx, args.seed ^ worker_idx, args.scale_factor, dataset_type=dataset_type) |
There was a problem hiding this comment.
Whoops - thanks for catching this. Instead of setting dataset_type here based on flags, I'd suggest we make it an argparse argument instead.
| parser.add_argument( | ||
| "--scale-factor", | ||
| type=int, | ||
| default=1, | ||
| default=6, |
There was a problem hiding this comment.
nit: Please keep this at 1 for now.
| if __name__ == "__main__": | ||
| yaml_main() | ||
| sys.exit(0) |
| parser.add_argument( | ||
| "--output-dir", | ||
| type=str, | ||
| default=".", | ||
| help="Environment variable that stores the output directory of tidb bench", | ||
| ) |
There was a problem hiding this comment.
Is this still used? Also it seems like the help string is not correct since the default is a path and not an environment variable?
amlatyrngom
force-pushed
the
aln-provisioning
branch
from
662f55c
to
07f634d
Compare
|
Closing since I think the intention was to not merge given the code divergence. |
No description provided.