Our doc regarding setting up Impala HA using HAProxy:
it uses the following timeout values:
Two issues here:
1. contimeout, clitimeout and srvtimeout are old, the new config names should be
The outdated config names shows that our docs are also outdated
2. the timeout value of 50000, which is 50 seconds is too low, it will cause impala client to timeout all the time, the correct value should be at least 1 hour to start with:
And we should also mention that these values should be dependant on how users use the cluster and how long their queries are running. It is NOT one config fits for all case.