Document the tpu-v6e-ci cluster used in E2E tests (#155)

Author: tengyifei

e2e_testing/README.md

@@ -21,5 +21,22 @@ model, and runs a few checks. This is implemented in
 - Check for specific log strings that indicate training success.
 - Check that there is a profile `.pb` file.
 
+## v6e XPK cluster
+
+E2E tests are launched onto an XPK cluster named `tpu-v6e-ci`.
+
+To heal or re-create the cluster, use the following:
+
+```sh
+xpk cluster create \
+    --tpu-type v6e-4 \
+    --cluster tpu-v6e-ci \
+    --num-slices 48 \
+    --on-demand \
+    --zone us-central2-b \
+    --project tpu-pytorch \
+    --default-pool-cpu-machine-type=n2-standard-32
+```
+
 [e2e-test]: /.github/workflows/e2e_test.yml
 [e2e-check]: /.github/workflows/reusable_e2e_check.yml