Fix AttributeError when client.get_default_node() returns None

[bacchuswng]

Pull Request check-list

Please make sure to review and check all of these items:

• Do tests and lints pass with this change?
• Do the CI tests pass with this change (enable it first in your forked repo and wait for the github action build to finish)?
• Is the new or changed code fully tested?
• Is a documentation update included (if this change modifies existing APIs, or introduces new ones)?
• Is there an example added to the examples folder (if applicable)?
• Was the change added to CHANGES file?

NOTE: these things are not required to open a PR and can be done
afterwards / while the PR is open.

Description of change

The default node is sometimes None. It's not obvious from the code how this happens (maybe an Exception that isn't being handled correctly, or a connection-related object is being closed while pipeline execution is still in progress). Regardless, we've observed AttributeError: 'NoneType' object has no attribute 'name' in practice.

Adding this defensive check ensures that we don't run into this issue.

[zs-neo]

Can you provide more information? How do you use redis-py? We are also trying related features and want to make sure we don’t encounter the same error.

[petyaslavova]

Hi @bacchuswng, can you please share more details about your code where you hit that issue - how do you use the RedisCluster client. I tried to reproduce something like it, but with no luck...
Before merging some fix, we should know how exactly we get into the problematic case and validate that this is the correct fix and not just hiding another problem...

[bacchuswng]

Hi @bacchuswng, can you please share more details about your code where you hit that issue - how do you use the RedisCluster client. I tried to reproduce something like it, but with no luck... Before merging some fix, we should know how exactly we get into the problematic case and validate that this is the correct fix and not just hiding another problem...

@petyaslavova Understood. Unfortunately, this specific issue is extremely hard to repro and only happened under load with a lot of concurrency.

Staring at the code, it is generally possible that client.get_default_node() returns None, but I haven't been able to find a smoking gun where, in this specific case, this call would return None. As a result, I view this PR as more a layer of defense that preserves the spirit of what the code is doing: replacing the default node, if there is one.