Make mget task claim strategy poll more frequently (elastic#190059)

In this PR, I'm changing the default `poll_interval` to `500` whenever
mget claim strategy is set. This keeps the `3000` value for the default
task claiming strategy and setting `xpack.task_manager.poll_interval`
still takes precedence.

I'm increasing the frequency because we no longer get as much contention
when adding Kibana nodes, allowing tasks to be picked up more frequently
/ on-time.

## To verify

1. Add a console log to the task manager plugin constructor (like
elastic@1e6b1ac)
2. Startup this PR without any task manager yml settings
3. Observe `3000` is the poll interval
4. Set `xpack.task_manager.claim_strategy: 'unsafe_mget'` in the
kibana.yml file and wait for Kibana restart
5. Observe `500` is the poll interval
6. Set `xpack.task_manager.poll_interval: 3001` and wait for Kibana
restart
7. Observe `3001` is the poll interval
8. Remove `xpack.task_manager.claim_strategy` setting and wait for
Kibana restart
9. Observe `3001` is still poll interval
10. Remove `xpack.task_manager.poll_interval` setting

Flaky test runner for mget tests:
https://buildkite.com/elastic/kibana-flaky-test-suite-runner/builds/6722

---------

Co-authored-by: Elastic Machine <[email protected]>

x-pack/plugins/task_manager/server/config.test.ts

@@ -5,7 +5,7 @@
  * 2.0.
  */
 
-import { configSchema } from './config';
+import { configSchema, CLAIM_STRATEGY_DEFAULT, CLAIM_STRATEGY_MGET } from './config';
 
 describe('config validation', () => {
   test('task manager defaults', () => {
@@ -242,4 +242,14 @@ describe('config validation', () => {
   test('any claim strategy is valid', () => {
     configSchema.validate({ claim_strategy: 'anything!' });
   });
+
+  test('default claim strategy defaults poll interval to 3000ms', () => {
+    const result = configSchema.validate({ claim_strategy: CLAIM_STRATEGY_DEFAULT });
+    expect(result.poll_interval).toEqual(3000);
+  });
+
+  test('mget claim strategy defaults poll interval to 500ms', () => {
+    const result = configSchema.validate({ claim_strategy: CLAIM_STRATEGY_MGET });
+    expect(result.poll_interval).toEqual(500);
+  });
 });

x-pack/plugins/task_manager/server/config.ts

@@ -13,6 +13,7 @@ export const MAX_CAPACITY = 50;
 export const MIN_CAPACITY = 5;
 export const DEFAULT_MAX_WORKERS = 10;
 export const DEFAULT_POLL_INTERVAL = 3000;
+export const MGET_DEFAULT_POLL_INTERVAL = 500;
 export const DEFAULT_VERSION_CONFLICT_THRESHOLD = 80;
 export const DEFAULT_MAX_EPHEMERAL_REQUEST_CAPACITY = MAX_WORKERS_LIMIT;
 
@@ -133,10 +134,18 @@ export const configSchema = schema.object(
       default: taskExecutionFailureThresholdSchema,
     }),
     /* How often, in milliseconds, the task manager will look for more work. */
-    poll_interval: schema.number({
-      defaultValue: DEFAULT_POLL_INTERVAL,
-      min: 100,
-    }),
+    poll_interval: schema.conditional(
+      schema.siblingRef('claim_strategy'),
+      CLAIM_STRATEGY_MGET,
+      schema.number({
+        defaultValue: MGET_DEFAULT_POLL_INTERVAL,
+        min: 100,
+      }),
+      schema.number({
+        defaultValue: DEFAULT_POLL_INTERVAL,
+        min: 100,
+      })
+    ),
     /* How many requests can Task Manager buffer before it rejects new requests. */
     request_capacity: schema.number({
       // a nice round contrived number, feel free to change as we learn how it behaves

x-pack/test/task_manager_claimer_mget/test_suites/task_manager/health_route.ts

@@ -129,7 +129,7 @@ export default function ({ getService }: FtrProviderContext) {
       const health = await getHealth();
       expect(health.status).to.eql('OK');
       expect(health.stats.configuration.value).to.eql({
-        poll_interval: 3000,
+        poll_interval: 500,
         monitored_aggregated_stats_refresh_rate: monitoredAggregatedStatsRefreshRate,
         monitored_stats_running_average_window: 50,
         monitored_task_execution_thresholds: {

x-pack/test/task_manager_claimer_mget/test_suites/task_manager/metrics_route.ts

@@ -52,7 +52,10 @@ export default function ({ getService }: FtrProviderContext) {
         // counters are reset every 30 seconds, so wait until the start of a
         // fresh counter cycle to make sure values are incrementing
         const initialMetrics = (
-          await getMetrics(false, (metrics) => metrics?.metrics?.task_claim?.value.total === 1)
+          await getMetrics(
+            false,
+            (metrics) => (metrics?.metrics?.task_claim?.value.total || 0) >= 1
+          )
         ).metrics;
         expect(initialMetrics).not.to.be(null);
         expect(initialMetrics?.task_claim).not.to.be(null);
@@ -92,7 +95,7 @@ export default function ({ getService }: FtrProviderContext) {
         const initialMetrics = (
           await getMetrics(
             false,
-            (metrics) => metrics?.metrics?.task_claim?.value.total === initialCounterValue
+            (metrics) => (metrics?.metrics?.task_claim?.value.total || 0) >= initialCounterValue
           )
         ).metrics;
         expect(initialMetrics).not.to.be(null);
@@ -101,7 +104,10 @@ export default function ({ getService }: FtrProviderContext) {
 
         // retry until counter value resets
         const resetMetrics = (
-          await getMetrics(false, (m: NodeMetrics) => m?.metrics?.task_claim?.value.total === 1)
+          await getMetrics(
+            false,
+            (m: NodeMetrics) => (m?.metrics?.task_claim?.value.total || 0) >= 1
+          )
         ).metrics;
         expect(resetMetrics).not.to.be(null);
         expect(resetMetrics?.task_claim).not.to.be(null);
@@ -113,7 +119,7 @@ export default function ({ getService }: FtrProviderContext) {
         const initialMetrics = (
           await getMetrics(
             false,
-            (metrics) => metrics?.metrics?.task_claim?.value.total === initialCounterValue
+            (metrics) => (metrics?.metrics?.task_claim?.value.total || 0) >= initialCounterValue
           )
         ).metrics;
         expect(initialMetrics).not.to.be(null);
@@ -133,8 +139,8 @@ export default function ({ getService }: FtrProviderContext) {
           expect(metrics?.task_claim).not.to.be(null);
           expect(metrics?.task_claim?.value).not.to.be(null);
 
-          expect(metrics?.task_claim?.value.success).to.equal(1);
-          expect(metrics?.task_claim?.value.total).to.equal(1);
+          expect(metrics?.task_claim?.value.success).to.be.greaterThan(0);
+          expect(metrics?.task_claim?.value.total).to.be.greaterThan(0);
 
           previousTaskClaimTimestamp = metrics?.task_claim?.timestamp!;