The failures of the majority of nodes will affect the majority of replicas in some log streams.
When the majority of nodes fail, some log streams will lack a leader for a long period. In this case, it is the top priority to recover services. You must identify the cause of the failure as soon as possible and try to troubleshoot the issue. For example, you can check whether the node failure is caused by a network or hardware exception. If you cannot troubleshoot the issue but the services cannot be affected for a long time, you can pull up a single replica. This is not a regular O&M operation but the last emergency action to take when the cluster cannot be recovered. It may cause data loss and two leaders. Proceed with caution under the guidance of OceanBase Technical Support. The details are not provided in this topic.