How to resolve "cassandra.cluster.NoHostAvailable" in a Python multi threaded program

Question

I am trying to insert records into Cassandra using a multi threaded Python program. I am running this program simultaneously on 3 machines. For some time being records are getting inserted, but later I am getting below exception. I am using driver provided by datastax.

cassandra.cluster.NoHostAvailable

I did some search and found out (source: https://datastax.github.io/python-driver/api/cassandra/cluster.html)

exception cassandra.cluster.NoHostAvailable
Raised when an operation is attempted but all connections are busy, defunct, closed, or resulted in errors when used.

My question is:
1. Is this a normal exception one could face with too many connection to Cassandra.
2. How would I resolve this in a situation where I want to create many connection/session to/of cassandra. (I know creating too many session is not advisable, it impact server performance as each session consume a handful amount of memory)

Below is the code fragment.

cluster = Cluster(['192.168.1.21'])
session = cluster.connect('myNameSpace')

def insertInToCassandra(catRange):
    for x in catRange:
        //function to insert records into Cassandra table

ProductRange = [
    range(900,920),
    range(921,940),
    range(941,960),
    range(961,980),
    range(981,1000)
     ]

# Make the Pool of workers
pool = ThreadPool(20)

# Open the urls in their own threads
# and return the results
results = pool.map(insertInToCassandra, ProductRange)

#close the pool and wait for the work to finish
pool.close()
pool.join()

Answer 1

That's a normal exception that may occur if one or more cassandra nodes is unavailable, especially if it goes into a GC spin or otherwise crashes.

Depending on your replication factor (RF) and consistency level (CL), a single node going offline may or may not break the application (with RF of 3 and CL of quorum, any individual node failing should be no problem).

You should check the health of your cassandra cluster with nodetool status, and check /var/log/cassandra/system.log for signs of nodes flapping up/down.