graphite-clickhouse doesn't close connection to clickhouse when request was canceled

[perederyaev]

Hi! I see a lot of long-running (up to 1 minute) requests from graphite-clickhouse (v0.13.4) in our clickhouse.

For reproducing the issue I make request to graphite-clickhouse with curl:

curl 'localhost:6000/render/?format=pickle&target= _________&from=1748882404&until=1749757204' -v -m2
*   Trying 127.0.0.1:6000...
* Connected to localhost (127.0.0.1) port 6000 (#0)
> GET /render/?format=pickle&target= _________&from=1748882404&until=1749757204 HTTP/1.1
> Host: localhost:6000
> User-Agent: curl/7.76.1
> Accept: */*
>
* Operation timed out after 2004 milliseconds with 0 bytes received
* Closing connection 0
curl: (28) Operation timed out after 2004 milliseconds with 0 bytes received

curl ends in 2 second and I see "context canceled" in the graphite-clickhouse log:

[2025-06-12T23:21:50.643+0300] INFO [render] query {"request_id": "968928b6d1eefe1cc2ce18b262bfa9a8", "query": "SELECT Path FROM graphite_index WHERE ((Level=4) AND (Path LIKE 'rb.rb%' AND match(Path, '_________'))) AND (Date >='2025-06-02' AND Date <= '2025-06-12') GROUP BY Path FORMAT TabSeparatedRaw", "total_rows_to_read": "7909864", "read_rows": "7909864", "read_bytes": "416941622", "written_rows": "0", "written_bytes": "0", "query_id": "968928b6d1eefe1cc2ce18b262bfa9a8::d9e0535b5257f227", "time": 0.132063994}
[2025-06-12T23:21:50.646+0300] INFO [render] finder {"request_id": "968928b6d1eefe1cc2ce18b262bfa9a8", "metrics": 6972, "find_cached": false}
[2025-06-12T23:21:52.513+0300] ERROR [render] data_parser {"request_id": "968928b6d1eefe1cc2ce18b262bfa9a8", "error": "context canceled", "read_bytes": 0, "runtime": "0s", "runtime_ns": 0}
[2025-06-12T23:21:52.513+0300] INFO [http] access {"request_id": "968928b6d1eefe1cc2ce18b262bfa9a8", "time": 2.002144606, "method": "GET", "url": "/render/?format=pickle&target=rb.rb*.rbmon.*&from=1748882404&until=1749757204", "peer": "[::1]:60226", "client": "", "status": 504, "find_cached": false}
[2025-06-12T23:21:57.802+0300] ERROR [render] reader {"request_id": "968928b6d1eefe1cc2ce18b262bfa9a8", "error": "context canceled"}

But query for graphite_reverse table in clickhouse continues to run up to 1 minute and ends with clickhouse error "Code: 24. DB::Exception: Cannot write to ostream at offset 18170421".

I've done some debugging and found that graphite-clickhouse doesn't close connection to clickhouse when request was canceled (curl finishes running) and waits for data-timeout (in my case, data-timeout = "1m0s") and only then disconnects from clickhouse.

That leads to many orphans queries in clickhouse and spare load on it.

Could you please fix the issue or provide any workaround?