[Fleet] Add retries w/ backoff to Fleet setup on Kibana boot

config/serverless.yml

@@ -7,6 +7,7 @@ xpack.fleet.internal.disableILMPolicies: true
 xpack.fleet.internal.disableProxies: true
 xpack.fleet.internal.activeAgentsSoftLimit: 25000
 xpack.fleet.internal.onlyAllowAgentUpgradeToKnownVersions: true
+xpack.fleet.internal.retrySetupOnBoot: true
 
 # Cloud links
 xpack.cloud.base_url: 'https://cloud.elastic.co'

package.json

@@ -887,6 +887,7 @@
     "email-addresses": "^5.0.0",
     "execa": "^5.1.1",
     "expiry-js": "0.1.7",
+    "exponential-backoff": "^3.1.1",
     "extract-zip": "^2.0.1",
     "fast-deep-equal": "^3.1.1",
     "fflate": "^0.6.9",

x-pack/plugins/fleet/common/types/index.ts

@@ -51,6 +51,7 @@ export interface FleetConfigType {
     fleetServerStandalone: boolean;
     onlyAllowAgentUpgradeToKnownVersions: boolean;
     activeAgentsSoftLimit?: number;
+    retrySetupOnBoot: boolean;
     registry: {
       kibanaVersionCheckEnabled: boolean;
       capabilities: string[];

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

@@ -188,6 +188,7 @@ export const config: PluginConfigDescriptor = {
               min: 0,
             })
           ),
+          retrySetupOnBoot: schema.boolean({ defaultValue: false }),
           registry: schema.object(
             {
               kibanaVersionCheckEnabled: schema.boolean({ defaultValue: true }),

x-pack/plugins/fleet/server/plugin.ts

@@ -5,6 +5,7 @@
  * 2.0.
  */
 
+import { backOff } from 'exponential-backoff';
 import type { Observable } from 'rxjs';
 import { BehaviorSubject } from 'rxjs';
 import { take, filter } from 'rxjs/operators';
@@ -532,18 +533,47 @@ export class FleetPlugin
           )
           .toPromise();
 
-        await setupFleet(
-          new SavedObjectsClient(core.savedObjects.createInternalRepository()),
-          core.elasticsearch.client.asInternalUser
+        // Retry Fleet setup w/ backoff
+        await backOff(
+          async () => {
+            await setupFleet(
+              new SavedObjectsClient(core.savedObjects.createInternalRepository()),
+              core.elasticsearch.client.asInternalUser
+            );
+          },
+          {
+            // We only retry when this feature flag is enabled
+            numOfAttempts: this.configInitialValue.internal?.retrySetupOnBoot ? Infinity : 1,
+            // 250ms initial backoff
+            startingDelay: 250,
+            // 5m max backoff
+            maxDelay: 60000 * 5,
+            timeMultiple: 2,
+            // avoid HA contention with other Kibana instances
+            jitter: 'full',
+            retry: (error: any, attemptCount: number) => {
+              const summary = `Fleet setup attempt ${attemptCount} failed, will retry after backoff`;
+              logger.debug(summary, { error: { message: error } });
+
+              this.fleetStatus$.next({
+                level: ServiceStatusLevels.available,
+                summary,
+                meta: {
+                  attemptCount,
+                  error,
+                },
+              });
+              return true;
+            },
+          }
         );
 
         this.fleetStatus$.next({
           level: ServiceStatusLevels.available,
           summary: 'Fleet is available',
         });
       } catch (error) {
-        logger.warn('Fleet setup failed');
-        logger.warn(error);
+        logger.warn('Fleet setup failed', { error: { message: error } });
 
         this.fleetStatus$.next({
           // As long as Fleet has a dependency on EPR, we can't reliably set Kibana status to `unavailable` here.

x-pack/plugins/fleet/server/services/epm/packages/_install_package.test.ts

@@ -135,6 +135,7 @@ describe('_installPackage', () => {
           disableProxies: false,
           fleetServerStandalone: false,
           onlyAllowAgentUpgradeToKnownVersions: false,
+          retrySetupOnBoot: false,
           registry: {
             kibanaVersionCheckEnabled: true,
             capabilities: [],
@@ -192,6 +193,7 @@ describe('_installPackage', () => {
           disableILMPolicies: false,
           fleetServerStandalone: false,
           onlyAllowAgentUpgradeToKnownVersions: false,
+          retrySetupOnBoot: false,
           registry: {
             kibanaVersionCheckEnabled: true,
             capabilities: [],
@@ -265,6 +267,7 @@ describe('_installPackage', () => {
             disableProxies: false,
             fleetServerStandalone: false,
             onlyAllowAgentUpgradeToKnownVersions: false,
+            retrySetupOnBoot: false,
             registry: {
               kibanaVersionCheckEnabled: true,
               capabilities: [],