[Doc] Update info for validate_after_inactivity option

[karenzone]

A link in the description of the validate_after_inactivity option is 404-ing. That's the danger of deep linking into other org's docs. OTOH, sending a user to the top level page with a "good luck" seems quite inadequate.

Here's the link that no longer resolves: https://hc.apache.org/httpcomponents-client-5.1.x/httpclient/apidocs/org/apache/http/impl/conn/PoolingHttpClientConnectionManager.html#setValidateAfterInactivity(int).

A direct replacement isn't readily obvious. We might be dealing with some EOL issues here?

[yaauie]

🤔 I think we should avoid the deep-link entirely, and rephrase the upstream guarantee in a way that is user-focused. A Logstash user who is trying to output to Elasticsearch should not need to understand the internals of a library we happen to use (much less one that is multiple layers deep, since this plugin uses Manticore, which in turn uses the HttpClient from Apache Commons).

It looks like a couple of things have happened with the upstream Apache Commons documentation, which I'm documenting here for reference

• httpclient version 5 has been renamed httpclient5
• In v5, this is now documented to take a TimeValue, not an int. I will have to look into the implementation to see if we really are using apache commons' http client v5 or an older version.
• the new "equivalent" link for v5 is
  https://hc.apache.org/httpcomponents-client-5.1.x/current/httpclient5/apidocs/org/apache/hc/client5/http/impl/io/PoolingHttpClientConnectionManager.html#setValidateAfterInactivity(org.apache.hc.core5.util.TimeValue),
• the "equivalent" link for 4.x is https://hc.apache.org/httpcomponents-client-4.5.x/current/httpclient/apidocs/org/apache/http/impl/conn/PoolingHttpClientConnectionManager.html#setValidateAfterInactivity(int)

---

I think we should rephrase this so that our users don't have to care so much about internals, maybe something like:

[id="plugins-{type}s-{plugin}-validate_after_inactivity"]
===== `validate_after_inactivity`

  * Value type is <<number,number>>
  * Default value is `10000`

How many milliseconds to wait before checking for a stale connection to determine if a keepalive request is needed.
Consider setting this value lower than the default, possibly to 0, if you get connection errors regularly.

This plugin maintains a pool of connections to Elasticsearch, which are independently kept alive and borrowed by workers.
Setting this to a lower value helps detect and refresh connections that have become stale (half-closed) _before_ they are given to a worker.
This reduces the amount of retries that are necessary when connections are volatile.

Setting this to a negative value disables pre-validation, requiring the worker to handle reconnection.