[BUG] BaseConnectionPool TypeError on 'http' ('publish_address' of undefined)

[jaco-smit]

What is the bug?

It looks like the property http is expected to be present at all times if you enable sniffing, but for some reason it is not there. My best guess is that it is due to the nodes not being available yet.
We are getting the following error:

“TypeError: Cannot read properties of undefined (reading ‘publish_address’)”,
   “ at ConnectionPool.nodesToHost (/opt/nodejs/node_modules/@opensearch-project/opensearch/lib/pool/BaseConnectionPool.js:232:31)”,“ at /opt/nodejs/node_modules/@opensearch-project/opensearch/lib/Transport.js:589:41”,
  “ at onBody (/opt/nodejs/node_modules/@opensearch-project/opensearch/lib/Transport.js:435:9)”,
  “ at IncomingMessage.onEnd (/opt/nodejs/node_modules/@opensearch-project/opensearch/lib/Transport.js:341:11)”

How can one reproduce the bug?

1. Connect to the existing cluster from the node js client
2. Onboard a new node to the cluster
3. After some seconds the node js client will throw an error and the script will crash

What is the expected behavior?

The undefined error should not happen

What is your host/environment?

Running this on AWS Opensearch 1.0 and also Elasticsearch 7.1.

Do you have any screenshots?

Logs pasted above.

Do you have any additional context?

This was fixed in the elasticsearch client here: elastic/elasticsearch-js@9856dfc

I checked the code and seems like this guardrail is not in place here. This should just be a copy of the fix they made.