add blobKeepAliveTimeout, queueKeepAliveTimeout, tableKeepAliveTimeout

[Slach]

fix #2053
workaround for ClickHouse/ClickHouse#60447

we use azurite for Azure BLOB storage in our integration tests
Azure CPP Sdk (or it usage on clickhouse-server side) don't respect Keep-Alive: 5 header which sent from azurite side, and continue to make request to already closed connection after 5 seconds, this header sent from standard nodejs http library

We are going to fix this behavior in the next version of clickhouse-server
but currently to allow pass e2e tests, we would like to propose adding cli option and config settings to allow increase value for Keep-Alive header

we tested these changes and it actually works, hope your CI/CD pipelines also will pass

[Slach]

@microsoft-github-policy-service agree

[blueww]

/azp run

[azure-pipelines]

Azure Pipelines could not run because the pipeline triggers exclude this branch/path.

[blueww]

@Slach

Thanks for the contribution!
Would you please rebase the code on the latest main to make PR validation can run?

And could you give more details on why the change can fix the issue?
And is the current default value aligned with product server behavior?
BTW, could you please also add test cases to make sure the new parameters really work?

[Slach]

rebased

[Slach]

PR description updated

[blueww]

@Slach

Besides test class "FSExtentStore" failure on BlobTest_Win_Node* is know issue which is in fixing, other test failures are new and related with this PR change.
Please help to fix the test failures.

Most of the new failure looks like :

  1) Blob Cors requests test
       Request Match rule exists for exact origin @loki @sql:
     FetchError: request to http://127.0.0.1:11000/devstoreaccount1?restype=service&comp=properties failed, reason: read ECONNRESET
      at ClientRequest.<anonymous> (node_modules/node-fetch/lib/index.js:1491:11)
      at ClientRequest.emit (node:events:519:28)
      at ClientRequest.emit (node:domain:488:12)
      at Socket.socketErrorListener (node:_http_client:500:9)
      at Socket.emit (node:events:519:28)
      at Socket.emit (node:domain:488:12)
      at emitErrorNT (node:internal/streams/destroy:169:8)
      at emitErrorCloseNT (node:internal/streams/destroy:128:3)
      at processTicksAndRejections (node:internal/process/task_queues:82:21)

[blueww]

What's the value before this PR?
Could we use the value before this PR is merged to avoid regression?

[Slach]

it was 5 by default

[blueww]

It looks might not be 5 for all scenarios, since so many test cases failed with "read ECONNRESET", looks the network break since keep live timeout.

[Slach]

sorry
i don't understand how to interpet stacktrace in #2443 (comment)

how to match which test exactly doesn't work? And how to run only this failed test?

i tried to run npm run test:blob locally on node 20

and got lot of failures can't figure out how to debug it

[blueww]

@Slach

If you click the "Details" of the failed jobs, you should can see the failure log, and find the failed test class/case name.

You might can first focus to make the following 2 classes pass with node 20 on win or Ubuntu:

• BlockBlobHighlevel
• ServiceAPIs

BTW, the failed test cases are all for "read ECONNRESET", it might caused by the keepalive timeout change might be too short.

[blueww]

@Slach

This RP looks target for a temp fix of CPP SDK issue (you said : "We are going to fix this behavior in the next version of clickhouse-server"), and this looks a high risk change since there are so many test cases failed.
So to avoid blocking other Azurite user, we might don't prefer to take this change.

I would suggest you contact CPP SDK team if this is a CPP SDK issue.
If you would like to use Azurite with the customized keeplivetimeout, I would suggest you build your own Azurite package from your code, instead push the change to Azurite public repo.

[blueww]

Close this PR as its looks a temp fix for client SDK issue, and take many Azurite Unit test fail.

[Slach]

@blueww
I figure out with tests, now it passed in my local machine could we reopen PR? or i shall create new?