[CONS-6832] Fix repetitive warning logs with log file tailer

[adel121]

What does this PR do?

This PR fixes an issue in autodiscovery component where container and kubelet listeners are subscribing to SourceAll in workloadmeta, which causes recreation of service that has been already created and finished.

As a result, the log agent is trying to make a new tailer for containers of cron jobs that have already stopped and are maintained on the node for some time in the Completed state depending on the user configuration.

This may or may not happen depending on the order the set, unset events are emitted by different collectors.

Until version 7.55, the kubelet listener was listening for events coming from the NodeOrchestrator collector (i.e. the kubelet in kubernetes). The container listener was listening for events coming from the Runtime collector (e.g. containerd runtime collector).

In 7.56, a fix was merged that solves an issue with duplicated logs.

In that fix, the workloadmeta subscription filters in both the kubelet and container listeners were changed to SourceAll (see here and here).

When a container is stopped, the containerd collector (or docker..) and the kubelet collector will both emit EventTypeUnset events.

If the subscriber had subscription with SourceAll, then if the kubelet unsets the container first, the event will be sent as a Set event to the subscriber because we still have another source (i.e. containerd collector) maintaining the entity in the store. The autodiscovery component will then receive a new container entity, but this time only populated with data coming from the container runtime and not from the kubelet. As a result, it will think that this should create a new service for this container because the data are different now.

This causes the log agent to try to make a file tailer. However, it won't be able to do so if the container has already been deleted fully from the node and the log file is removed. It will thus emit an extra warning saying that it failed to make a file tailer.

Example:

Could not make file tailer for source container_collect_all (falling back to socket): cannot find pod for container "6f585560fac9d45127f20509c7c84d017126776573772f42e1bd45af59090e54": "6f585560fac9d45127f20509c7c84d017126776573772f42e1bd45af59090e54" not found

Motivation

Fix CONS-6832

Describe how you validated your changes

We need to verify 2 things:

By restoring the sources of the kubelet and container listeners to NodeOrchestrator and Runtime, we would still need to
test that previous fix still works. We need to repeat the same QA as the previous fix.

Already tested it with a cronjob, and didn't get any duplicate logs for the job:

You should also verify that we almost never get a warning saying Could not make file tailer for source in the agent logs.

(Note: you should set the log level to warning when doing the QA, and you should leave it running for a sufficient time, preferably an hour at least, before validating that you are not getting many warnings)

Possible Drawbacks / Trade-offs

None

Additional Notes

[adel121]

It is not clear to me why subscription filters were changed to SourceAll in the previous fix.

I already verified that the previous fix still works even if this is kept as SourceRuntime and NodeOrchestrator.

@L3n41c don't hesitate to add your thoughts if you have more context about this.

In the mean time, I will proceed with this PR as the freeze is close, it fixes a customer issue and doesn't seem to break your fix.

[clamoriniere]

I don't remember the issue, But I think @L3n41c tried in the past to change this source but it has some side effect.

the one that I'm thinking of (but not 100%sure) is when user don't give access to user to the container runtime, but we can still have info about the container with the kubelet.

[adel121]

I see your point, but I think the kubelet listener already generates container service when it receives events from the kubelet source. (see here)

Keeping the container listener filter source as SourceAll and changing the kubelet's listener filter source back to NodeOrchestrator seems to also work (i.e. no duplicate logs, no extra warnings).

I think that since the kubelet listener is listening for events for NodeOrchestrator, then it also includes events from the kubelet about the containers. So it should still work IMO even if the container listener is using events from the container runtime and we don't have access to the runtime.

WDYT?

[agent-platform-auto-pr]

Test changes on VM

Use this command from test-infra-definitions to manually test this PR changes on a VM:

inv aws.create-vm --pipeline-id=51849344 --os-family=ubuntu

Note: This applies to commit 8b2f373

[agent-platform-auto-pr]

Uncompressed package size comparison

Comparison with ancestor 70abd234c3950faf793ae73b1276bd6d1dc85c46

Diff per package

package	diff	status	size	ancestor	threshold
datadog-agent-amd64-deb	0.00MB	✅	1191.10MB	1191.10MB	140.00MB
datadog-iot-agent-aarch64-rpm	0.00MB	✅	108.88MB	108.88MB	10.00MB
datadog-agent-x86_64-rpm	0.00MB	✅	1200.42MB	1200.42MB	140.00MB
datadog-agent-x86_64-suse	0.00MB	✅	1200.42MB	1200.42MB	140.00MB
datadog-agent-aarch64-rpm	0.00MB	✅	944.69MB	944.69MB	140.00MB
datadog-agent-arm64-deb	0.00MB	✅	935.39MB	935.39MB	140.00MB
datadog-dogstatsd-amd64-deb	0.00MB	✅	78.57MB	78.57MB	10.00MB
datadog-dogstatsd-x86_64-rpm	0.00MB	✅	78.64MB	78.64MB	10.00MB
datadog-dogstatsd-x86_64-suse	0.00MB	✅	78.64MB	78.64MB	10.00MB
datadog-dogstatsd-arm64-deb	0.00MB	✅	55.77MB	55.77MB	10.00MB
datadog-heroku-agent-amd64-deb	0.00MB	✅	505.27MB	505.27MB	70.00MB
datadog-iot-agent-amd64-deb	0.00MB	✅	113.35MB	113.35MB	10.00MB
datadog-iot-agent-arm64-deb	0.00MB	✅	108.81MB	108.81MB	10.00MB
datadog-iot-agent-x86_64-rpm	-0.00MB	✅	113.42MB	113.42MB	10.00MB
datadog-iot-agent-x86_64-suse	-0.00MB	✅	113.42MB	113.42MB	10.00MB

Decision

✅ Passed

[cit-pr-commenter]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 5c30cd54-50fa-4e41-931c-9b306a66ab51

Baseline: 70abd23
Comparison: 8b2f373
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	tcp_syslog_to_blackhole	ingress throughput	+1.42	[+1.36, +1.48]	1	Logs
➖	file_to_blackhole_1000ms_latency_linear_load	egress throughput	+0.68	[+0.21, +1.15]	1	Logs
➖	quality_gate_idle_all_features	memory utilization	+0.62	[+0.54, +0.70]	1	Logs bounds checks dashboard
➖	quality_gate_idle	memory utilization	+0.18	[+0.14, +0.21]	1	Logs bounds checks dashboard
➖	file_to_blackhole_0ms_latency	egress throughput	+0.12	[-0.70, +0.94]	1	Logs
➖	tcp_dd_logs_filter_exclude	ingress throughput	+0.00	[-0.01, +0.01]	1	Logs
➖	file_to_blackhole_300ms_latency	egress throughput	-0.00	[-0.64, +0.64]	1	Logs
➖	file_to_blackhole_100ms_latency	egress throughput	-0.00	[-0.65, +0.65]	1	Logs
➖	uds_dogstatsd_to_api	ingress throughput	-0.01	[-0.12, +0.11]	1	Logs
➖	file_to_blackhole_0ms_latency_http2	egress throughput	-0.08	[-0.96, +0.80]	1	Logs
➖	file_to_blackhole_0ms_latency_http1	egress throughput	-0.13	[-1.06, +0.80]	1	Logs
➖	file_to_blackhole_1000ms_latency	egress throughput	-0.17	[-0.98, +0.64]	1	Logs
➖	file_to_blackhole_500ms_latency	egress throughput	-0.23	[-1.00, +0.54]	1	Logs
➖	file_tree	memory utilization	-0.24	[-0.37, -0.12]	1	Logs
➖	uds_dogstatsd_to_api_cpu	% cpu utilization	-1.38	[-2.06, -0.70]	1	Logs
➖	quality_gate_logs	% cpu utilization	-2.99	[-6.19, +0.20]	1	Logs

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	links
✅	file_to_blackhole_0ms_latency	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency	memory_usage	10/10
✅	file_to_blackhole_0ms_latency_http1	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency_http1	memory_usage	10/10
✅	file_to_blackhole_0ms_latency_http2	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency_http2	memory_usage	10/10
✅	file_to_blackhole_1000ms_latency	memory_usage	10/10
✅	file_to_blackhole_1000ms_latency_linear_load	memory_usage	10/10
✅	file_to_blackhole_100ms_latency	lost_bytes	10/10
✅	file_to_blackhole_100ms_latency	memory_usage	10/10
✅	file_to_blackhole_300ms_latency	lost_bytes	10/10
✅	file_to_blackhole_300ms_latency	memory_usage	10/10
✅	file_to_blackhole_500ms_latency	lost_bytes	10/10
✅	file_to_blackhole_500ms_latency	memory_usage	10/10
✅	quality_gate_idle	memory_usage	10/10	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	bounds checks dashboard
✅	quality_gate_logs	lost_bytes	10/10
✅	quality_gate_logs	memory_usage	10/10

Explanation

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

CI Pass/Fail Decision

✅ Passed. All Quality Gates passed.

• quality_gate_idle_all_features, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check lost_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.