Query receive timeout always uses default setting

[ShivaS123]

ClickHouse server version 19.4.1.1

Whenever set receive_timeout is used (e.g. set receive_timeout=6000), the change is reflected in system.settings:

SELECT *
FROM system.settings
WHERE name IN ('max_execution_time', 'receive_timeout')
 
┌─name───────────────┬─value─┬─changed─┬─description─┐
│ receive_timeout    │ 6000  │       1 │             │
│ max_execution_time │ 0     │       0 │             │
└────────────────────┴───────┴─────────┴─────────────┘

However, a long query (e.g. optimize...final) always receives an error after 300 seconds:

Timeout exceeded while receiving data from server. Waited for 300 seconds, timeout is 300 seconds.
Cancelling query.

The only way to increase the receive_timeout is to connect with clickhouse-client --receive_timeout=6000

from processlist:

Settings.Names:       ['max_threads','connect_timeout_with_failover_ms','receive_timeout','send_timeout','use_uncompressed_cache','load_balancing','distributed_aggregation_memory_efficient','max_bytes_before_external_group_by','max_bytes_before_external_sort','max_memory_usage']
Settings.Values:      ['12','10000','6000','6000','0','random','1','30000000000','30000000000','60000000000']

(haven't checked other timeouts)

Thanks!

[alex-zaitsev]

Could be related to #4215

[den-crane]

I think this parameter is applied to CH server, when CH server is a client itself and executes a query against another server.

And it's not obvious how client should/can use this. Because if ch-client pulls value of this parameter from a server and use it then CH server also have to pull this when it plays as a client?

I think receive_timeout needs a better documentation. But it works as designed.