cqlsh is unable to start on slow/loaded clusters

[vladzcloudius]

Setup HEAD: branch-1.6 01b0514098b37c28efe7ef2ce290552e48ed30c2

Description cqlsh is missing a way to configure a cassandra.cluster.Cluster()'s control_connection_timeout parameter. By default this value is set to 2s, which has proven to be not enough. As a result on slow or loaded Nodes this results in cqlsh failure to start.

$ cqlsh
Connection error: ('Unable to connect to any servers', {'127.0.0.1': OperationTimedOut('errors=None, last_host=None',)})

This issue is a duplicate of C*'s cqlsh issue: https://issues.apache.org/jira/browse/CASSANDRA-10959

We need to:

1. Increase a default.
2. Make sure cqlsh allows to override the default by either/both value in the cqlshrc or/and the command line option.

[vladzcloudius]

The issue looks addressed on the master branch.

[penberg]

@vladzcloudius Do you know which commit fixed this issue?

[vladzcloudius]

@penberg I don't but the cqlsh in enterprise is very different from the one in the master:

• In the enterprise there is only cqlsh.
• In the master cqlsh calls the cqlsh.py and the contents on cqlsh.py is very different from the one in the cqlsh in the enterprise.

Another thing - I'm not sure master doesn't have this issue. I didn't test it and we definitely have to check this.

[vladzcloudius]

@penberg @slivne @nirmaayan Just to clarify - the customer in question has received the patched version that doesn't have a proper fix but rather hard codes the big timeout values. The proper fix should be to make cqlsh take the user provided parameters into an account like described in the issue.

[vladzcloudius]

@slivne @nirmaayan Guys, is this issue still relevant? Have we verified that the cqlsh in our latest releases doesn't have this issue? We need some QA on this and if it's not relevant anymore - close this issue.