New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Creating per session cloud table on read requests #1232
Comments
Do you mean the StorageCloudMergeTree object in the worker? Different sub-query from server may search for different parts in the same table.(addDataParts, loadDataParts). So it should not be used outside the query session. |
Yes I mean StorageCloudMergeTree object in the worker. I understand loading different parts, but recreating table every time creates an interesting error on retry or fallback when a timeout or rpc socket connection issue occurs: 2024.02.29 07:15:03.325072 [ 1494590 ] {3ba693d9-d6d2-11ee-892c-ba3d820a74a5} (448054137376211845) TCPHandler: Code: 57, e.displayText() = DB::Exception: DB::Exception: Table ed_mt_v1.ed_metric_2d6be233-f7bb-4fe1-90a5-28a95c86ec9c_448054137376211845 already exists. SQLSTATE: 42P07.: While executing Remote SQLSTATE: 42P07, Stack trace:
|
@smmsmm1988 Could you check about the exception and confirm whether it is a knowable issue? |
@yuanzhubi No, it seems to be a new issue. @luffyhwl Could you help check this exception? |
@ozcelgozde commit ac18e6a
|
We are actually running the latest master and still observe the same error sometimes. When it happens, I also observe this error from server:
From read workers: 2024.03.04 06:34:40.540763 [ 28754 ] {} PlanSegmentExecutor: [420c95fe-d9f1-11ee-a2dd-a6d5e601c8a5_1]: Query has excpetion with code: 2010, detail
|
Any number of possible errors return table already exists :) |
Question
I was looking into some timeouts during reads and noticed we are creating a cloud table over and over again for each read transaction even though table has not changed at all. This causes 3-10 seconds delay for each query and sometimes timeouts. I was wondering why read replicas dont keep a copy of the queried tables once its loaded until an modify event occurs in which case, they can reload?
The text was updated successfully, but these errors were encountered: