Disaster recovery in Managed ClickHouse®
Managed ClickHouse® prevents and mitigates emergencies or crises with multiple disaster recovery methods to keep your data safe and sound.
Disaster recovery is a process of coping with emergencies or crises using dedicated methods for protecting resources and/or reestablishing their desired status. In the context of data infrastructure, well-established disaster recovery methods are of a particular importance for preventing data loss or corruption. Software failure, loss of a node, or datacenter outage are only a few examples of emergencies when disaster recovery comes in.
High availability
High availability (HA) is an entity’s ability to continuously maintain a certain level of operational performance for a desired period of time. HA is typically achieved by redundancy - securing replicas of databases or services to be highly available. To support disaster recovery technologies, a database service needs to stay highly available, for example, by operating on a few nodes holding the same data.
With Exoscale, HA for your service is supported in Business and Premium plans. See Plan comparison for details.
Backup and restore
Service backup
Backups of Managed ClickHouse services happen automatically on a daily basis.
Backups cover the following:
- Access entities (for example, users, roles, passwords, or secrets) stored in ZooKeeper
- Database definitions
- Table schemas
- Table content (
part files) - Dictionaries
You can restore your service from a selected backup.
Note
Part files With the ClickHouse’s ReplicatedMergeTree table engine, each INSERT query results in creating a new file, so-called part, written only once and not modifiable.
Using part files allows incremental backups in Managed ClickHouse: only changed parts are backed up and files already available in the object storage are left out from the backup.
For more information on backups in Exoscale, see Backups at Exoscale.
Service recovery
You can create a copy of your Managed ClickHouse service and restore the data from a selected service backup. For this purpose, you create a fork from the original service. This spins up a new service that hosts the data recovered from the selected backup.
Sharding
Essentially, sharding is a technique of splitting database rows across multiple database nodes, which usually significantly increases performance. Each shard is in turn replicated across several nodes, so the rows it holds survive the loss of any single node. Replication at the shard level is what provides high availability and helps to achieve disaster recovery.
With Managed ClickHouse Business and Premium plans, each shard is replicated across three nodes running on separate hosts in the zone. The service and the data stay available even if a node is lost.
Note
Although sharding with replicated nodes can reduce failures, it still cannot save a service from the loss of an entire zone.
For information on how to work with shards in Managed ClickHouse, see Enable reading and writing data across shards.
Limitations
Managed ClickHouse has a few restrictions on the disaster recovery capability.
- No backup to another zone
- No point in time recovery (PITR)
For all the restrictions and limits for Managed ClickHouse, see Managed ClickHouse limits and limitations.