BulkCopy not raising exception on big datasets

[marcinatpayholdingcom]

BulkCopy not raising exception on big datasets

[DarkWanderer]

Hi
Thank you for your report, that would be a serious bug. Can you clarify a bit more on how this appeared - was the exception happening somewhere near the beginning of dataset or near the end?

Also, if you would be able to reproduce it - does the issue reproduce on previous versions, e.g. 6.8.1?

[DarkWanderer]

I have tried to reproduce the issue in a test and it doesn't seem to appear with a synthetic test - a serialization exception is always correctly propagated. Can you provide more context please?

[marcinatpayholdingcom]

I will try to build demo for you. On 9 May 2024, at 12:11, Oleg V. Kozlyuk ***@***.***> wrote:  I have tried to reproduce the issue in a test and it doesn't seem to appear with a synthetic test - a serialization exception is always correctly propagated. Can you provide more context please? — Reply to this email directly, view it on GitHub<#475 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AYPRFDYB2YAGH2KUWALCNJTZBNDUBAVCNFSM6AAAAABHL2EAQSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBSGM3DMMZTG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[DarkWanderer]

Please try version 7.5.0 to see if it fixes the issue for you

[marcinatpayholdingcom]

Nop, still not working correctly, here is a demo to replicate this problem. Let me know, maybe I am doing something wrong.
https://github.com/marcinatpayholdingcom/clickhouseclient_bug_replication

[DarkWanderer]

Sorry - took me a while to get my hands on it. However, for me it works perfectly:

Hello, World!

O:\Projects\clickhouseclient_bug_replication\ClickHouse.ClientNotSerializing\bin\Debug\net8.0\ClickHouse.ClientNotSerializing.exe (process 21308) exited with code 0.
Press any key to close this window . . .

Any additional details I am missing? Do you use any special flags or parameters (e.g. UseSessions)? What OS was this reproed on?

Here are specific exceptions which are shown in the debugger:

[marcinatpayholdingcom]

I am running this on Windows
this part should throw the serialization exception but it does not:

I do not have any special setting, I just run it as it is in the project.
So the problem is, it does not throw, so only part of the data is inserted and chunks with broken data are skipped.

[DarkWanderer]

For me it is triggered quite perfectly every time

Just to double check, when you say "exception is not thrown", have you checked it in the debugger by putting a breakpoint on the logger._LogError line, or did you check only by looking at the logs? Because your test project does not output any logs even in case of error

[marcinatpayholdingcom]

by setting the brekpoin at the log line and by seeing what data got inserted into Clickhouse.
When you get the serialization exception that means that no data got inserted into Clickhouse in the first attempt, am I right?

In my case there is some data inserted.

I think you can close this issue as not able to replicate. I have no clue what setup causes this situation on my machine.
First I thought that it was because of the original project I am working on, maybe I did something wrong. But when I created this demo and things are still failing incorrectly then I give up.

[DarkWanderer]

I think there may be just a misunderstanding here, which means I'll need to make docs more clear

When you get the serialization exception that means that no data got inserted into Clickhouse in the first attempt, am I right?

ClickHouse is not transactional (other than experimental features) - and hence BulkCopy adapter isn't as well. If you have a large set of data and the serialization (or insertion) has failed for one of the batches, it is expected that some of the batches may have made it through already, and it falls on the higher-level code to handles the situation. I.e. in this particular case, 20k of rows get into database before exception - and that's by design.

If you need to have some semblance of atomicity, the data needs to go as one 'batch' - so you can try setting BatchSize to Int32.MaxValue (reference). It would be slow and memory-consuming, but you would get a binary result.