[CI] Cloud purge: improve logging, add CLOUD_DELETE_ON_ERROR (elastic…

…#208392)

## Summary
When not run through the official way, cloud deployments can get weird
names (e.g.: `kibana-pr-false` - from incorrect parameterization through
testing:
https://buildkite.com/elastic/kibana-migration-pipeline-staging/builds/192).

This throws a wrench in the gears for the purge script. Although the
removals still work, the script exits with status code 1, which fails
the pipeline.

This PR adds some more resilience to the script, better logging of
problematic cases, and a flag to allow removal of problematic cases (if
someone manually runs the pipeline with `CLOUD_DELETE_ON_ERROR=1`).

Tested: 
- only fail and log:
https://buildkite.com/elastic/kibana-purge-cloud-deployments/builds/6710
- CLOUD_DELETE_ON_ERROR=1, to remove problematic case
https://buildkite.com/elastic/kibana-purge-cloud-deployments/builds/6712
(fails for some reason on `keep_...` instance, but never again?)
- With the `+e`, and exit code comparison
https://buildkite.com/elastic/kibana-purge-cloud-deployments/builds/6713

.buildkite/scripts/steps/cloud/purge_deployments.sh

@@ -1,9 +1,17 @@
 #!/usr/bin/env bash
 
-set -euo pipefail
+set -uo pipefail
+# Note, -e is not set above, so we can capture all errors, and attempt purge individually
 
 echo '--- Purging Cloud deployments'
 ts-node .buildkite/scripts/steps/cloud/purge_deployments.ts
+EXIT_CODE_CLOUD=$?
 
 echo '--- Purging Project deployments'
 ts-node .buildkite/scripts/steps/cloud/purge_projects.ts
+EXIT_CODE_PROJECTS=$?
+
+if [ $EXIT_CODE_CLOUD -ne 0 ] || [ $EXIT_CODE_PROJECTS -ne 0 ]; then
+  echo "❌ Purge failed (EXIT_CODE_CLOUD=$EXIT_CODE_CLOUD | EXIT_CODE_PROJECTS=$EXIT_CODE_PROJECTS)"
+  exit 1
+fi

.buildkite/scripts/steps/cloud/purge_deployments.ts

@@ -11,20 +11,27 @@ import { execSync } from 'child_process';
 import { getKibanaDir } from '#pipeline-utils';
 
 const deploymentsListJson = execSync('ecctl deployment list --output json').toString();
-const { deployments } = JSON.parse(deploymentsListJson);
+const { deployments } = JSON.parse(deploymentsListJson) as {
+  deployments: Array<{ name: string; id: string }>;
+};
 
-const prDeployments = deployments.filter((deployment: any) =>
-  deployment.name.startsWith('kibana-pr-')
-);
+const prDeployments = deployments.filter((deployment) => deployment.name.startsWith('kibana-pr-'));
 
-const deploymentsToPurge = [];
+const deploymentsToPurge: typeof deployments = [];
 
 const NOW = new Date().getTime() / 1000;
 const DAY_IN_SECONDS = 60 * 60 * 24;
+const CLOUD_DELETE_ON_ERROR = process.env.CLOUD_DELETE_ON_ERROR?.match(/^(true|1)$/i);
 
 for (const deployment of prDeployments) {
   try {
-    const prNumber = deployment.name.match(/^kibana-pr-([0-9]+)$/)[1];
+    const prNumber: string | undefined = deployment.name.match(/^kibana-pr-([0-9]+)$/)?.[1];
+    if (!prNumber) {
+      throw new Error(
+        `Invalid deployment name: ${deployment.name}; expected kibana-pr-{PR_NUMBER}).`
+      );
+    }
+
     const prJson = execSync(`gh pr view '${prNumber}' --json state,labels,updatedAt`).toString();
     const pullRequest = JSON.parse(prJson);
     const prOpen = pullRequest.state === 'OPEN';
@@ -59,9 +66,13 @@ for (const deployment of prDeployments) {
       deploymentsToPurge.push(deployment);
     }
   } catch (ex) {
+    console.error(`Error deleting deployment (${deployment.id}; ${deployment.name})`);
     console.error(ex.toString());
-    // deploymentsToPurge.push(deployment); // TODO should we delete on error?
-    process.exitCode = 1;
+    if (CLOUD_DELETE_ON_ERROR) {
+      deploymentsToPurge.push(deployment);
+    } else {
+      process.exitCode = 1;
+    }
   }
 }