Best practise to load balance the database-query instances? #3900
Replies: 6 comments 2 replies
-
I think it also relevant to query fallback retry and we should also be able to fetch query result consistently from any of the instance in a cluster.
|
Beta Was this translation helpful? Give feedback.
-
For the cloud, there is no |
Beta Was this translation helpful? Give feedback.
-
fair enough, we can weak the instance-wise system tables in the distributed environment. |
Beta Was this translation helpful? Give feedback.
-
This is helpful for killing query without switching databend-query nodes(may be helpful for non cloud users)? for example:
|
Beta Was this translation helpful? Give feedback.
-
Hello, I encounter load balancing problems when using HTTP Handler. If a large amount of data is returned, data will be returned in paging mode. When the load balancing policy is |
Beta Was this translation helpful? Give feedback.
-
Currently, we do not have definitive plans to support these solutions in future updates. However, we highly recommend configuring your Router to direct the same request (with same query id) always to one server. |
Beta Was this translation helpful? Give feedback.
-
currently databend-query have some in-memory states, like query logs, metrics, etc (mostly are the in memory system tables?). a SQL fired on a different databend-query instance may get a different result, like:
if I want run databend-query instances in a HA manner with an Round Robin load balancer, The proxy can not produce a consistent result between queries except enable a sticky session.
also, as the system administrator, i hope to manage the cluster running state in a global manner, if someone produces a nuclear bomb query, i can KILL it in time to save the entire cluster. but currently i need execute SHOW PROCESSLIST on every instances to find which instance the SQL is fired.
any suggestion about this?
Beta Was this translation helpful? Give feedback.
All reactions