Cloud: add page for Make Before Break (MBB)

docs/cloud/features/04_infrastructure/automatic_scaling.md → docs/cloud/features/04_infrastructure/automatic_scaling/01_auto_scaling.md

@@ -38,10 +38,15 @@ For Enterprise tier services scaling works as follows:
   - Custom profiles (`highMemory` and `highCPU`) do not support vertical autoscaling or manual vertical scaling. However, these services can be scaled vertically by contacting support.
 
 :::note
-Scaling in ClickHouse Cloud happens in what we call "Make Before Break" (MBB) approach. This adds one or more replicas of the new size before removing the old replicas, preventing any loss of capacity during scaling operations. By eliminating the gap between removing existing replicas and adding new ones, MBB creates a more seamless and less disruptive scaling process. It is especially beneficial in scale-up scenarios, where high resource utilization triggers the need for additional capacity, since removing replicas prematurely would only exacerbate the resource constraints. As part of this approach we wait up to an hour to let any existing queries complete on the older replicas before we will remove them. This balances the need for existing queries to complete, while at the same time ensuring that older replicas do not linger around for too long.
+Scaling in ClickHouse Cloud happens in what we call a ["Make Before Break" (MBB)](/cloud/features/mbb) approach.
+This adds one or more replicas of the new size before removing the old replicas, preventing any loss of capacity during scaling operations.
+By eliminating the gap between removing existing replicas and adding new ones, MBB creates a more seamless and less disruptive scaling process.
+It is especially beneficial in scale-up scenarios, where high resource utilization triggers the need for additional capacity, since removing replicas prematurely would only exacerbate the resource constraints.
+As part of this approach, we wait up to an hour to let any existing queries complete on the older replicas before removing them.
+This balances the need for existing queries to complete, while at the same time ensuring that older replicas do not linger around for too long.
 
 Please note that as part of this change: 
-1. Historical system table data will be retained for up to a maximum of 30 days as part of scaling events. In addition, any system table data older than December 19, 2024, for services on AWS or GCP and older than January 14, 2025, for services on Azure will not be retained as part of the migration to the new organization tiers.
+1. Historical system table data is retained for up to a maximum of 30 days as part of scaling events. In addition, any system table data older than December 19, 2024, for services on AWS or GCP and older than January 14, 2025, for services on Azure will not be retained as part of the migration to the new organization tiers.
 2. For services utilizing TDE (Transparent Data Encryption) system table data is currently not maintained after MBB operations. We are working on removing this limitation.
 :::
 

docs/cloud/features/04_infrastructure/automatic_scaling/02_make_before_break.md

@@ -0,0 +1,41 @@
+---
+sidebar_position: 1
+sidebar_label: 'Make Before Break (MBB)'
+slug: /cloud/features/mbb
+description: 'Page describing Make Before Break (MBB) operations in ClickHouse Cloud'
+keywords: ['Make Before Break', 'MBB', 'Scaling', 'ClickHouse Cloud']
+title: 'Make Before Break (MBB) operations in ClickHouse Cloud'
+doc_type: 'guide'
+---
+
+import Image from '@theme/IdealImage';
+import mbb_diagram from '@site/static/images/cloud/features/mbb/vertical_scaling.png';
+
+ClickHouse Cloud performs cluster upgrades and cluster scaling utilizing a **Make Before Break** (MBB) approach.
+In this approach, new replicas are added to the cluster before removing old replicas from it.
+This is as opposed to the break-first approach, where old replicas would first be removed, before adding new ones.
+
+The MBB approach has several benefits:
+* Since capacity is added to the cluster before removal, the **overall cluster capacity does not go down** unlike with the break-first approach. Of course, unplanned events such as node or disk failures etc. can still happen in a cloud environment.
+* This approach is especially useful in situations where the cluster is under heavy load as it **prevents existing replicas from being overloaded** as would happen with a break-first approach.
+* Because replicas can be added quickly without having to wait to remove replicas first, this approach leads to a **faster, more responsive** scaling experience.
+
+The image below shows how this might happen for a cluster with 3 replicas where the service is scaled vertically:
+
+<Image img={mbb_diagram} size="lg" alt="Example diagram for a cluster with 3 replicas which gets vertically scaled" />
+
+Overall, MBB leads to a seamless, less disruptive scaling and upgrade experience compared to the break-first approach previously utilized.
+
+With MBB, there are some key behaviors that users need to be aware of:
+
+1. MBB operations wait for existing workloads to finish on the current replicas before being terminated.
+   This period is currently set to 1 hour, which means that scaling or upgrades can wait up to one hour for a long-running query on a replica before the replica is removed.
+   Additionally, if a backup process is running on a replica, it is left to complete before the replica is terminated.
+2. Due to the fact that there is a waiting time before a replica is terminated, there can be situations where a cluster might have more than the maximum number of replicas set for the cluster.
+   For example, you might have a service with 6 total replicas, but with an MBB operation in progress, 3 additional replicas may get added to the cluster leading to a total of 9 replicas, while the older replicas are still serving queries. 
+   This means that for a period of time, the cluster will have more than the desired number of replicas. 
+   Additionally, multiple MBB operations themselves can overlap, leading to replica accumulation. This can happen, for instance, in scenarios where several vertical scaling requests are sent to the cluster via the API.
+   ClickHouse Cloud has checks in place to restrict the number of replicas that a cluster might accumulate.
+3. With MBB operations, system table data is kept for 30 days. This means every time an MBB operation happens on a cluster, 30 days worth of system table data is replicated from the old replicas to the new ones.
+
+If you are interested in learning more about the mechanics of MBB operations, please look at this [blog post](https://clickhouse.com/blog/make-before-break-faster-scaling-mechanics-for-clickhouse-cloud) from the ClickHouse engineering team.

docs/cloud/features/04_infrastructure/automatic_scaling/_category_.json

@@ -0,0 +1,5 @@
+{
+  "label": "Automatic Scaling",
+  "collapsible": true,
+  "collapsed": true,
+}