Repo sync

data/release-notes/enterprise-server/3-13/13.yml

@@ -16,6 +16,8 @@ sections:
       For instances in a high availability configuration, Elasticsearch indices were deleted on failover and when `ghe-repl-teardown REPLICA_HOSTNAME` was run from the primary instance. All indices are recoverable except audit log indices, whose source of truth is Elasticsearch itself.
     - |
       In Azure environments, running `ghe-single-config-apply` or `ghe-repl-setup` resulted in "Permission denied" errors during the pre-flight check.
+    - |
+      For appliances in a high availability configuration, Elasticsearch indices were deleted either on failover, or when running `ghe-repl-teardown <REPLICA_HOSTNAME>` from the primary instance.
   changes:
     - |
       Elasticsearch shards are excluded from the replica node when stopping replication via `ghe-repl-stop`. To prevent Elasticsearch from being stopped before all shards have been removed, Elasticsearch is polled until the shard count on the replica node is zero instead of waiting for a maximum timeout of 30 seconds.

data/release-notes/enterprise-server/3-14/10.yml

@@ -18,6 +18,8 @@ sections:
       In Azure environments, running `ghe-single-config-apply` or `ghe-repl-setup` resulted in "Permission denied" errors during the pre-flight check.
     - |
       On instances with a GitHub Advanced Security license, some secret scanning alerts were opened incorrectly despite the relevant folders or files being excluded from secret scanning.
+    - |
+      For appliances in a high availability configuration, Elasticsearch indices were deleted either on failover, or when running `ghe-repl-teardown <REPLICA_HOSTNAME>` from the primary instance.
   changes:
     - |
       Elasticsearch shards are excluded from the replica node when stopping replication via `ghe-repl-stop`. To prevent Elasticsearch from being stopped before all shards have been removed, Elasticsearch is polled until the shard count on the replica node is zero instead of waiting for a maximum timeout of 30 seconds.

data/release-notes/enterprise-server/3-15/5.yml

@@ -14,6 +14,8 @@ sections:
       For instances in a high availability configuration, Elasticsearch indices were deleted on failover and when `ghe-repl-teardown REPLICA_HOSTNAME` was run from the primary instance. All indices are recoverable except audit log indices, whose source of truth is Elasticsearch itself.
     - |
       On instances with a GitHub Advanced Security license, some secret scanning alerts were opened incorrectly despite the relevant folders or files being excluded from secret scanning.
+    - |
+      For appliances in a high availability configuration, Elasticsearch indices were deleted either on failover, or when running `ghe-repl-teardown <REPLICA_HOSTNAME>` from the primary instance.
   changes:
     - |
       Elasticsearch shards are excluded from the replica node when stopping replication via `ghe-repl-stop`. To prevent Elasticsearch from being stopped before all shards have been removed, Elasticsearch is polled until the shard count on the replica node is zero instead of waiting for a maximum timeout of 30 seconds.

data/release-notes/enterprise-server/3-16/1.yml

@@ -20,6 +20,8 @@ sections:
       Domain entries could fail to load in the "Verified & Approves Domains" section of the site admin dashboard if one or more authoritative nameservers for the affected domain was unreachable or unresponsive due to inefficient DNS queries.
     - |
       On instances with a GitHub Advanced Security license, some secret scanning alerts were opened incorrectly despite the relevant folders or files being excluded from secret scanning.
+    - |
+      For appliances in a high availability configuration, Elasticsearch indices were deleted either on failover, or when running `ghe-repl-teardown <REPLICA_HOSTNAME>` from the primary instance.
   changes:
     - |
       Elasticsearch shards are excluded from the replica node when stopping replication via `ghe-repl-stop`. To prevent Elasticsearch from being stopped before all shards have been removed, Elasticsearch is polled until the shard count on the replica node is zero instead of waiting for a maximum timeout of 30 seconds.