Add lower bound on poll_interval

[tawAsh1]

#39163

Add lower bound on poll_interval

$ curl -H "Content-Type: application/json" -X PUT -d '{
  "persistent": {
    "indices.lifecycle.poll_interval": "1ms",
    "logger.org.elasticsearch.xpack.indexlifecycle": "TRACE"
  }
}' localhost:9200/_cluster/settings

{"error":{"root_cause":[{"type":"illegal_argument_exception","reason":"failed to parse value [1ms] for setting [indices.lifecycle.poll_interval], must be >= [1s]"}],"type":"illegal_argument_exception","reason":"failed to parse value [1ms] for setting [indices.lifecycle.poll_interval], must be >= [1s]"},"status":400}

[elasticmachine]

Pinging @elastic/es-core-features

[gwbrown]

@elasticmachine test this please

I think we may have to adjust some of the places we set this setting in our tests as well, so this test run may fail.

[tawAsh1]

There are tests which is related to this setting.
All tests set poll_interval to 1 sec or more.
Therefore, this change probably does not affect the test results.

elasticsearch/x-pack/plugin/ilm/src/test/java/org/elasticsearch/xpack/indexlifecycle/IndexLifecycleInitialisationTests.java

Lines 404 to 427 in d79cd8c

	public void testPollIntervalUpdate() throws Exception {
	TimeValue pollInterval = TimeValue.timeValueSeconds(randomLongBetween(1, 5));
	final String server_1 = internalCluster().startMasterOnlyNode(
	Settings.builder().put(LifecycleSettings.LIFECYCLE_POLL_INTERVAL, pollInterval.getStringRep()).build());
	IndexLifecycleService indexLifecycleService = internalCluster().getInstance(IndexLifecycleService.class, server_1);
	assertBusy(() -> {
	assertNotNull(indexLifecycleService.getScheduler());
	assertThat(indexLifecycleService.getScheduler().jobCount(), equalTo(1));
	});
	{
	TimeValueSchedule schedule = (TimeValueSchedule) indexLifecycleService.getScheduledJob().getSchedule();
	assertThat(schedule.getInterval(), equalTo(pollInterval));
	}
	// update the poll interval
	TimeValue newPollInterval = TimeValue.timeValueHours(randomLongBetween(6, 1000));
	Settings newIntervalSettings = Settings.builder().put(LifecycleSettings.LIFECYCLE_POLL_INTERVAL,
	newPollInterval.getStringRep()).build();
	assertAcked(client().admin().cluster().prepareUpdateSettings().setTransientSettings(newIntervalSettings));
	{
	TimeValueSchedule schedule = (TimeValueSchedule) indexLifecycleService.getScheduledJob().getSchedule();
	assertThat(schedule.getInterval(), equalTo(newPollInterval));
	}
	}

elasticsearch/x-pack/plugin/ilm/src/test/java/org/elasticsearch/xpack/indexlifecycle/IndexLifecycleServiceTests.java

Lines 92 to 95 in ecf6af4

	Settings settings = Settings.builder().put(LifecycleSettings.LIFECYCLE_POLL_INTERVAL, "1s").build();
	when(clusterService.getClusterSettings()).thenReturn(new ClusterSettings(settings,
	Collections.singleton(LifecycleSettings.LIFECYCLE_POLL_INTERVAL_SETTING)));
	when(clusterService.lifecycleState()).thenReturn(State.STARTED);

[gwbrown]

You're right! I was actually thinking of things like this:

elasticsearch/x-pack/plugin/ilm/qa/multi-cluster/build.gradle

Line 25 in 89b80c6

setting 'indices.lifecycle.poll_interval', '1000ms'

Where the setting is defined in the Gradle setup for our tests - I thought some of them used poll intervals of <1s, but I was wrong. Thanks for your contribution @tawAsh1!