fix(ARC, networking): improve request settings

[dsmith3197]

This does a few things:

• Reworks the structs so that component-specific defaults are propagated to their docs, similar to batch settings.
• Changes default_retry_max_duration_secs from 1 hour to 30 seconds. 1 hour is way too high for this value. If a downstream goes down, this means that Vector might wait up to an hour before it retries after the downstream service recovers. This will significantly improve Vector's ability to quickly recover after a downstream service failure.
• Fixes a few bugs where we incorrectly applied defaults to TowerRequestConfig. Particularly, we previously (1) ignored component-level default overrides, using the global default instead and (2) ignored the user-set concurrency value if it equaled the global default (e.g., if a component overrode the default concurrency to be none and the user specified adaptive, the resulting concurrency setting would be none).
• Adds a max_concurrency_limit option to AdaptiveConcurrencySettings. This provides documentation to the user what the maximum is and allows them to override it as they desire. The lack of visibility here has raised questions in the past.

https://datadoghq.atlassian.net/browse/OPW-157

[datadog-vectordotdev]

Datadog Report

Branch report: dougsmith/update-request-settings
Commit report: 56e4028

✅ vector: 0 Failed, 0 New Flaky, 2007 Passed, 0 Skipped, 1m 20.58s Wall Time

[pront]

We could reduce even more e.g. 30 seconds.

[dsmith3197]

Yeah I'll change it. I'm in favor of being more aggressive here. Retries are relatively cheap even at a 30 second interval.

[dsmith3197]

Is this a change we actually want to make? It wasn't being applied previously.

[jszwedko]

Agreed, this change is significant. I would argue for keeping it out for now. I think we need to the ability to route failed events from sinks before we lower the retry count such that it may start dropping data.

[lukesteensen]

Agree with keeping it out. Do we have context on why it should be 5?

[dsmith3197]

Opted to remove this and use the new global default value of 30 seconds.

[dsmith3197]

This is the global default value so I removed it.

[dsmith3197]

This is the global default value so I removed it.

[dsmith3197]

@lukesteensen I did a significant refactoring of this PR to match how we configure batch settings so that the component-specific overrides are propagated to their reference docs.

While doing that, I began questioning whether or not the existing default overrides are sensible. They weren't being applied previously... so this may introduce significant changes to the default behavior for particular sinks.

[lukesteensen]

I'm curious why these sinks have such a lower retry count.

[dsmith3197]

I suspect it was added to one of the metric sinks and then was copied to the others.

[lukesteensen]

Agree with keeping it out. Do we have context on why it should be 5?

[dsmith3197]

Update on the component-level default overrides

As of today, there are components that override the following config settings:

• Concurrency
• Timeout
• Rate limit
• Retry attempts

Our proposal is the following:

• Concurrency
  • Action: Leave as is. This results in no behavior changes. The sinks that limit concurrency do it for a specific reason.
• Timeout
  • Current overrides
    • Cloudwatch & SQS lower the timeout to 30 seconds
  • Action: Remove. There wasn't a rationale given when these were introduced.
• Rate limit
  • Current overrides
    • Cloudwatch metrics rate limits to 150
    • Azure Blob rate limits to 250
    • Chronicle Unstructured, GCS & Stackdriver rate limit to 1000
  • Action: Leave as is. Most, if not all, of these were added by reading the documentation of the given service.
• Retry attempts
  • Current overrides
    • Datadog Metrics, Datadog Traces, InfluxDB Logs & Metrics, Sematext Metrics limits retries to 5 attempts
  • Action: Remove. Retry attempts are limited to 5 for several metrics sinks, but that limit was not previously being applied. I propose we remove these overrides to retain the existing behavior, as we should be cautious about limiting the number of request retries.

[jszwedko]

It seems like there are some default behavior changes here. Do we need/want to add some notes to the upgrade guide?

[jszwedko]

Awesome, thanks!

[github-actions]

Regression Detector Results

Run ID: 0c510cb2-fd14-4384-b940-3c7385a7436d
Baseline: 52068d4
Comparison: 7dc7d2e
Total vector CPUs: 7

Explanation

A regression test is an integrated performance test for vector in a repeatable rig, with varying configuration for vector. What follows is a statistical summary of a brief vector run for each configuration across SHAs given above. The goal of these tests are to determine quickly if vector performance is changed and to what degree by a pull request.

Because a target's optimization goal performance in each experiment will vary somewhat each time it is run, we can only estimate mean differences in optimization goal relative to the baseline target. We express these differences as a percentage change relative to the baseline target, denoted "Δ mean %". These estimates are made to a precision that balances accuracy and cost control. We represent this precision as a 90.00% confidence interval denoted "Δ mean % CI": there is a 90.00% chance that the true value of "Δ mean %" is in that interval.

We decide whether a change in performance is a "regression" -- a change worth investigating further -- if both of the following two criteria are true:

1. The estimated |Δ mean %| ≥ 5.00%. This criterion intends to answer the question "Does the estimated change in mean optimization goal performance have a meaningful impact on your customers?". We assume that when |Δ mean %| < 5.00%, the impact on your customers is not meaningful. We also assume that a performance change in optimization goal is worth investigating whether it is an increase or decrease, so long as the magnitude of the change is sufficiently large.

2. Zero is not in the 90.00% confidence interval "Δ mean % CI" about "Δ mean %". This statement is equivalent to saying that there is at least a 90.00% chance that the mean difference in optimization goal is not zero. This criterion intends to answer the question, "Is there a statistically significant difference in mean optimization goal performance?". It also means there is no more than a 10.00% chance this criterion reports a statistically significant difference when the true difference in mean optimization goal is zero -- a "false positive". We assume you are willing to accept a 10.00% chance of inaccurately detecting a change in performance when no true difference exists.

The table below, if present, lists those experiments that have experienced a statistically significant change in mean optimization goal performance between baseline and comparison SHAs with 90.00% confidence OR have been detected as newly erratic. Negative values of "Δ mean %" mean that baseline is faster, whereas positive values of "Δ mean %" mean that comparison is faster. Results that do not exhibit more than a ±5.00% change in their mean optimization goal are discarded. An experiment is erratic if its coefficient of variation is greater than 0.1. The abbreviated table will be omitted if no interesting change is observed.

No interesting changes in experiment optimization goals with confidence ≥ 90.00% and |Δ mean %| ≥ 5.00%.

Fine details of change detection per experiment.

experiment	goal	Δ mean %	Δ mean % CI	confidence
datadog_agent_remap_blackhole_acks	ingress throughput	+1.10	[+1.01, +1.19]	100.00%
http_elasticsearch	ingress throughput	+1.00	[+0.94, +1.06]	100.00%
file_to_blackhole	egress throughput	+0.96	[-1.38, +3.29]	49.91%
splunk_hec_route_s3	ingress throughput	+0.90	[+0.39, +1.41]	99.65%
datadog_agent_remap_datadog_logs_acks	ingress throughput	+0.74	[+0.66, +0.81]	100.00%
datadog_agent_remap_blackhole	ingress throughput	+0.59	[+0.50, +0.67]	100.00%
fluent_elasticsearch	ingress throughput	+0.41	[-0.05, +0.86]	86.09%
syslog_loki	ingress throughput	+0.31	[+0.26, +0.35]	100.00%
syslog_log2metric_humio_metrics	ingress throughput	+0.23	[+0.12, +0.34]	99.91%
http_to_http_acks	ingress throughput	+0.17	[-1.14, +1.49]	17.30%
http_to_http_noack	ingress throughput	+0.10	[+0.00, +0.20]	91.41%
http_to_http_json	ingress throughput	+0.03	[-0.04, +0.10]	48.22%
splunk_hec_indexer_ack_blackhole	ingress throughput	+0.00	[-0.14, +0.14]	1.81%
splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.00	[-0.14, +0.14]	0.24%
http_to_s3	ingress throughput	-0.02	[-0.29, +0.26]	7.24%
otlp_grpc_to_blackhole	ingress throughput	-0.03	[-0.13, +0.07]	37.83%
splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.05	[-0.17, +0.06]	57.25%
http_text_to_http_json	ingress throughput	-0.12	[-0.25, -0.00]	90.01%
enterprise_http_to_http	ingress throughput	-0.13	[-0.21, -0.05]	99.26%
datadog_agent_remap_datadog_logs	ingress throughput	-0.25	[-0.34, -0.15]	100.00%
syslog_splunk_hec_logs	ingress throughput	-0.60	[-0.64, -0.56]	100.00%
syslog_regex_logs2metric_ddmetrics	ingress throughput	-0.72	[-0.81, -0.62]	100.00%
syslog_log2metric_splunk_hec_metrics	ingress throughput	-1.15	[-1.28, -1.01]	100.00%
syslog_humio_logs	ingress throughput	-1.27	[-1.33, -1.20]	100.00%
socket_to_socket_blackhole	ingress throughput	-2.29	[-2.37, -2.22]	100.00%
otlp_http_to_blackhole	ingress throughput	-3.14	[-3.28, -3.00]	100.00%