[lbarrosby]

What this PR does / why we need it:

Follow-up to #2113, as suggested in this review. The Fluentd elasticsearch and opensearch ClusterOutput templates hardcode logstashFormat: true, so setting fluentd.output.es.logstashFormat or fluentd.output.opensearch.logstashFormat to false has no effect.

Both templates now use dig "logstashFormat" true ., same as #2113: the default stays true when the key is not set, and an explicit false is kept.

Checked with helm template (Fluentd with both outputs enabled):

• key not set: logstashFormat: true, full render identical to master
• logstashFormat=true: true
• logstashFormat=false: true on master, false with this change

helm lint charts/fluent-operator passes, and the rendered ClusterOutputs validate with kubeconform against the chart's fluentd.fluent.io_clusteroutputs CRD schema.

Which issue(s) this PR fixes:

None, follow-up from the #2113 review.

Does this PR introduced a user-facing change?

Helm chart: fluentd.output.es.logstashFormat and fluentd.output.opensearch.logstashFormat can now be set to false.

Additional documentation, usage docs, etc.:

[joshuabaird]

Thanks, this looks good. I rendered both outputs with the key unset and with logstashFormat=false and got true and false as expected.

One small suggestion: add a commented # logstashFormat: true under fluentd.output.es in values.yaml, like the Fluent Bit ES section has, so the option is easier to find.

Two existing issues I noticed while testing, not caused by this PR:

• fluentd-output-opensearch.yaml uses .buffer.type and .buffer.path inside {{- with .buffer }}. Setting any OpenSearch buffer value fails to render with nil pointer evaluating interface {}.type. They should be .type and .path.
• Both Fluentd templates ignore buffer.enable, so the buffer block renders even with enable: false, which is the ES default in values.yaml.

Happy to see those handled in a follow-up.

[lbarrosby]

Added it under fluentd.output.es, thanks.