Is there a proxy solution for Yugabyte Cassandra?

[Steve_Liang]

I see there is some talk about proxy solution for YEDIS:

Achieving real zero-downtime YEDIS by using a proxy

Given a 3 node setup that all run YEDIS, you can use any of the TServers as backends for clients. But when you have to update one node, you have to restart it, causing downtime (or it might go down for any other reason). If you’re too lazy to change clients to only talk to the 2 online nodes, you can use a proxy that takes care of this automatically. Common choices are twemproxy by Twitter and dynomit by Netflix. Each TServer runs 1 instance of either twemproxy or dynomite. The config file for…

There is also proxy solution for Postgres:

Using Envoy Proxy's PostgreSQL & TCP Filters to Collect Yugabyte SQL...

We are looking for proxy solution for Yugabyte YCQL (Cassandra). Does YB support proxy against YCQL?

[dorian_yugabyte]

Hi @Steve_Liang

YCQL drivers are cluster aware and have no need for a proxy. They know which nodes are up/down and can route queries directly. YCQL also supports multi-plexed connections to run concurrent queries over 1 connection.

What language will you be coding in ?

[Steve_Liang]

We have developers using Java, Node.JS, and Python. My primary concern is the number of connections the master node handles. May be asking about proxy for YB cluster is the wrong question.

[dorian_yugabyte]

@Steve_Liang

We have developers using Java, Node.JS, and Python.

These will work fine.

@Steve_Liang

My primary concern is the number of connections the master node handles.

The clients don’t connect to the yb-master processes but yb-tservers.
Do you have a use case where you instantly start 1000 client processes ?

[Steve_Liang]

Dorian, we run an increasing number of K8S apps and at times one app can scale to hundreds of pods. I seen on occasions where we scale 100 pods to 200 pods in one go. In production operation, we found that YB team asked us to reduce the number of connection to YB cluster which composes of 6 m5.4xlarge nodes. I am guessing it comes down to efficient use of YB client driver and beyond that, create another YB cluster.

[dorian_yugabyte]

Yes, the problem is that when making a new connection to yb-tserver, yb-tserver then queries yb-master, which can get overwhelmed.

We are in the process of caching queries to yb-master to significantly lower overhead. See [YCQL] Make connection creation cheaper and master-less · Issue #4640 · yugabyte/yugabyte-db · GitHub and [ycql] Investigate/address expensive "USE keyspace;" queries · Issue #5872 · yugabyte/yugabyte-db · GitHub.

[Steve_Liang]

Sounds good. Looks like a fundamental change in YB.