[Synthetics] Detect agent-level issues in monitor integration health

[miguelmartin-elastic]

Blocked by #256738

Summary

Closes #258541. Follow-up to #256738.

Extends the monitor integration health API and UI to detect agent-level issues in private locations — specifically when no agents are enrolled or all agents are offline/unhealthy. These statuses are surfaced alongside the existing integration-level checks.

Monitors Management - No agents enrolled tooltip
Monitors Management - Mixed issues: reset-fixable and non-reset-fixable
Monitors Management: reset is applied only to monitors that have at least one reset-fixable issue
Monitors Management: reset is applied only to monitors that have at least one reset-fixable issue
Monitor edit: if there are no reset-fixable issues the reset button is not shown
Private locations table: if none of the unhealthy monitors is reset-fixable in that private location, the reset button is not shown
Private locations table: if at least one of the unhealthy monitors is reset-fixable in that private location, the reset button is shown

New health statuses (server):

• missing_agents — the agent policy exists but has zero active agents enrolled
• unhealthy_agent — agents are enrolled but none are online

Agent status is fetched in batch via Fleet's getAgentStatusForAgentPolicy for all relevant agent policies. The check uses status.active (not status.all) so that unenrolled/deleted agents don't incorrectly count as enrolled.

Priority order (most fundamental → least):
missing_location → missing_agent_policy → missing_package_policy → missing_agents → unhealthy_agent → healthy

UI changes:

• MissingAgentPolicy, MissingAgents, and UnhealthyAgent are classified as non-reset-fixable — the reset button is hidden when all unhealthy locations have agent-level issues
• When a selection is mixed (some fixable, some not), the bulk reset modal shows a collapsible warning listing the skipped monitors
• In the private locations table, the reset button is now a primary inline action to avoid a blank space when hidden
• The edit monitor callout shows "Agent issue detected" as the title when agent-level issues are present

Reset API fix:
Before calling editMonitors, the reset API now pre-filters out locations whose agent policy no longer exists. This prevents AgentPolicyNotFoundError from bubbling up as a 500 when a monitor has both fixable and non-fixable locations.

Test plan

Prerequisites

• A running Kibana with at least one private location that has an enrolled, online agent

Setup test monitors

Run ~/elastic/scripts/break_monitors.sh against your Kibana instance. It creates:

Monitor	Locations	Expected status
Mon A	loc1 (agent online)	missing_package_policy (fixable)
Mon B	loc1 (agent online)	missing_package_policy (fixable)
Mon C	loc1 + loc2 (no agents)	missing_package_policy on loc1 + missing_agents on loc2
Mon D	loc2 (no agents)	missing_agents (not fixable)
Mon E	loc3 (deleted agent policy)	missing_agent_policy (not fixable)
Mon F	loc1 + loc3	missing_package_policy on loc1 + missing_agent_policy on loc3

What to verify

Monitor list page (/app/synthetics/monitors):

• Mon A and Mon B show the unhealthy badge and a "Reset monitor" button in the row actions
• Mon D shows the unhealthy badge but no reset button
• Mon C shows the unhealthy badge and a reset button (mixed: one location is fixable)
• Selecting Mon A + Mon B + Mon D and clicking bulk reset opens the confirmation modal with a warning listing Mon D as skipped
• Confirming the bulk reset fixes Mon A and Mon B (they become healthy after a few seconds)

Edit monitor page for Mon C:

• The callout lists both locations with their respective status messages
• The reset button is visible (because loc1 is fixable)
• Clicking reset fixes loc1; loc2 remains missing_agents

Private locations settings page (/app/synthetics/settings/private-locations):

• The location with enrolled agents shows a "Reset monitors" button when Mon A/B are broken
• The no-agents location does not show the reset button (all issues are agent-level)
• No blank space appears where the reset button would be on the no-agents location row

New health status messages:

• missing_agents: "No Fleet agents are enrolled in the agent policy for this private location. Enroll an agent in Fleet to resolve this."
• unhealthy_agent: "All Fleet agents for this private location are unhealthy or offline. Check the agent status in Fleet."

[macroscopeapp]

Approvability

Verdict: Would Approve

Adds new health detection capability for Fleet agent issues in Synthetics monitoring. While this introduces new runtime behavior (new health statuses and Fleet API calls), all changes are within the author's owned code, are additive/backward-compatible, and include comprehensive tests. The scope is self-contained within the health API.

No code changes detected at 174e407. Prior analysis still applies.

^{Macroscope would have approved this PR. A repo admin can enable approvability here.}

[elasticmachine]

💚 Build Succeeded

• Buildkite Build
• Commit: 174e407

Metrics [docs]

Async chunks

Total size of all lazy-loaded chunks that will be downloaded as the user navigates the app

id	before	after	diff
synthetics	1.1MB	1.1MB	+516.0B

History

• 💛 Build #424561 was flaky 151ffee
• 💛 Build #423908 was flaky a263c62

[shahzad31]

LGTM !!

I think we need to do further work regarding making these errors bit more prominent in the UI

we likely also need to display these on the cards and also likely as top level item in case it's happening for a significant amount of monitors