[Heartbeat] Add changelog/docs for new scheduler (#14891)

We added a new scheduler in fa63445 but forgot to add the docs / changelog. This fixes that.
elastic · Dec 2, 2019 · 01b3862 · 01b3862
1 parent fa63445
commit 01b3862
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 0 deletions.
diff --git a/CHANGELOG.next.asciidoc b/CHANGELOG.next.asciidoc
@@ -54,6 +54,7 @@ https://github.com/elastic/beats/compare/v7.0.0-alpha2...master[Check the HEAD d
 - Removed `host.name` field that should never have been included. Heartbeat uses `observer.*` fields instead. {pull}14140[14140]
 - Changed default user-agent to be `Elastic-Heartbeat/VERSION (PLATFORM_INFO)` as the current default `Go-http-client/1.1` is often blacklisted. {pull}14291[14291]
 - JSON/Regex checks against HTTP bodies will only consider the first 100MiB of the HTTP body to prevent excessive memory usage. {pull}14223[pull]
+- Heartbeat now starts monitors scheduled with the '@every X' syntax instantaneously on startup, rather than waiting for the given interval to pass before running them. {pull}14890[14890]
 
 *Journalbeat*
 
@@ -216,6 +217,7 @@ https://github.com/elastic/beats/compare/v7.0.0-alpha2...master[Check the HEAD d
 - Fix integer comparison on JSON responses. {pull}13348[13348]
 - Fix storage of HTTP bodies to work when JSON/Regex body checks are enabled. {pull}14223[14223]
 - Fix recording of SSL cert metadata for Expired/Unvalidated x509 certs. {pull}13687[13687]
+- The heartbeat scheduler no longer drops scheduled items when under very high load causing missed deadlines. {pull}14890[14890]
 
 *Journalbeat*
 

diff --git a/heartbeat/docs/heartbeat-options.asciidoc b/heartbeat/docs/heartbeat-options.asciidoc
@@ -139,6 +139,13 @@ started.
 The `schedule` option uses a cron-like syntax based on https://github.com/gorhill/cronexpr#implementation[this `cronexpr` implementation],
 but adds the `@every` keyword.
 
+For stats on the execution of scheduled tasks you can enable the HTTP stats server with `http.enabled: true` in heartbeat.yml, then run `curl http://localhost:5066/stats | jq .heartbeat.scheduler` to view the scheduler's stats. Stats are provided for both jobs and tasks. Each time a monitor is scheduled is considered to be a single job, while portions of the work a job does, like DNS lookups and executing network requests are defined as tasks. The stats provided are:
+
+* **jobs.active:** The number of actively running jobs/monitors.
+* **jobs.missed_deadline:** The number of jobs that executed after their scheduled time. This can be caused either by overlong long timeouts from the previous job or high load preventing heartbeat from keeping up with work.
+* **tasks.active:** The number of tasks currently running.
+* **tasks.waiting:** If the global `schedule.limit` option is set, this number will reflect the number of tasks that are ready to execute, but have not been started in order to prevent exceeding `schedule.limit`.
+
 [float]
 [[monitor-ipv4]]
 ==== `ipv4`