[Security Solution] fix flaky endpoint ftr tests (#152119)

x-pack/plugins/fleet/server/services/epm/elasticsearch/retry.test.ts

@@ -85,4 +85,13 @@ describe('retryTransientErrors', () => {
     await expect(retryTransientEsErrors(esCallMock)).rejects.toThrow(error);
     expect(esCallMock).toHaveBeenCalledTimes(1);
   });
+
+  it('retries with additionalResponseStatuses', async () => {
+    const error = new EsErrors.ResponseError({ statusCode: 123, meta: {} as any, warnings: [] });
+    const esCallMock = jest.fn().mockRejectedValueOnce(error).mockResolvedValue('success');
+    expect(await retryTransientEsErrors(esCallMock, { additionalResponseStatuses: [123] })).toEqual(
+      'success'
+    );
+    expect(esCallMock).toHaveBeenCalledTimes(2);
+  });
 });

x-pack/plugins/fleet/server/services/epm/elasticsearch/retry.ts

@@ -17,24 +17,29 @@ const retryResponseStatuses = [
   410, // Gone
 ];
 
-const isRetryableError = (e: any) =>
+const isRetryableError = (e: any, additionalResponseStatuses: number[] = []) =>
   e instanceof EsErrors.NoLivingConnectionsError ||
   e instanceof EsErrors.ConnectionError ||
   e instanceof EsErrors.TimeoutError ||
-  (e instanceof EsErrors.ResponseError && retryResponseStatuses.includes(e?.statusCode!));
+  (e instanceof EsErrors.ResponseError &&
+    [...retryResponseStatuses, ...additionalResponseStatuses].includes(e?.statusCode!));
 
 /**
  * Retries any transient network or configuration issues encountered from Elasticsearch with an exponential backoff.
  * Should only be used to wrap operations that are idempotent and can be safely executed more than once.
  */
 export const retryTransientEsErrors = async <T>(
   esCall: () => Promise<T>,
-  { logger, attempt = 0 }: { logger?: Logger; attempt?: number } = {}
+  {
+    logger,
+    attempt = 0,
+    additionalResponseStatuses = [],
+  }: { logger?: Logger; attempt?: number; additionalResponseStatuses?: number[] } = {}
 ): Promise<T> => {
   try {
     return await esCall();
   } catch (e) {
-    if (attempt < MAX_ATTEMPTS && isRetryableError(e)) {
+    if (attempt < MAX_ATTEMPTS && isRetryableError(e, additionalResponseStatuses)) {
       const retryCount = attempt + 1;
       const retryDelaySec = Math.min(Math.pow(2, retryCount), 64); // 2s, 4s, 8s, 16s, 32s, 64s, 64s, 64s ...
 

x-pack/plugins/fleet/server/services/epm/elasticsearch/transform/install.ts

@@ -703,9 +703,13 @@ async function handleTransformInstall({
   // start transform by default if not set in yml file
   // else, respect the setting
   if (startTransform === undefined || startTransform === true) {
-    await esClient.transform.startTransform(
-      { transform_id: transform.installationName },
-      { ignore: [409] }
+    await retryTransientEsErrors(
+      () =>
+        esClient.transform.startTransform(
+          { transform_id: transform.installationName },
+          { ignore: [409] }
+        ),
+      { logger, additionalResponseStatuses: [400] }
     );
     logger.debug(`Started transform: ${transform.installationName}`);
   }

x-pack/test/security_solution_endpoint/apps/endpoint/index.ts

@@ -15,8 +15,7 @@ import {
 export default function (providerContext: FtrProviderContext) {
   const { loadTestFile, getService } = providerContext;
 
-  // FLAKY: https://github.com/elastic/kibana/issues/72874
-  describe.skip('endpoint', function () {
+  describe('endpoint', function () {
     const ingestManager = getService('ingestManager');
     const log = getService('log');
     const endpointTestResources = getService('endpointTestResources');