High availability cluster (HA) to build [Translation]

xiaoxiao2021-03-06  109

?

Highly available cluster construction

Solution 1: The Mon HeartBeat Fake Coda Solution

MON: A universal resource monitoring system that monitors the availability and service nodes of network services.

HeartBeat: Provides a heartbeat detection between two nodes through "serial line" or "UDP".

FAKE: The host "takeover" of the IP layer is realized by ARP spoof.

• The server's fault recovery is processed: "Mon" background program runs on the load balancer, which is responsible for monitoring the "service process" and "service node" in the cluster. FPING.MONITOR Each T second will take the initiative to detect if the service node is still alive; other related service monitor actively detects the service process on each node every m second. For example: http.monitor can detect HTTP services; FTP.Monitor is used to detect FTP services; When the "Services Node" or "Service Process" is dead or resurrected, the alert will automatically remove or add a rule in Linux Vertual Server Table. Therefore, the load balancer will automatically block the failure of the "service process" or "service node", and when they recover, the service is automatically resumed.

??? Now, the load balancer has become a single failure point of the entire system. In order to block the failure of the load balancer (main load balancer), we need to set up a backup load balancer. "Fake" software can be used in this backup machine (from the load balancer), when the primary load balancer is invalid, from the load balancer can take over the IP address of the primary load balancer to provide services. "HeartBeat" is used to detect the state of the primary load balancer, thereby deciding whether "fake" from the load balancer is enabled.

But, whether it is "failed recovery" or "service takeover" between the master-slave server, it will cause the currently established connection failure, which requires the client to resend their request. ?

??? Network file system CODA is a fault-tolerant distributed file system that is the successor of the Andrew file system. Introducing a network file system, mainly in order to resolve the storage consistency encountered in the cluster.

Solution 2: The LDIRECTORD HeartBeat SOLUTION

??? LDIRECTORD (Linux Director Daemon) is a separate process that monitors the service of real servers. Currently, only two services are supported: HTTP and HTTPS. It only needs to work with HeartBeat.

??? LDIRECTORD compared to MON's strengths: ??? 1) It is specifically written for LVS ??????? It reads the configuration file /etc/ha.d/xxx.cf, which contains All IPvs routing table configurations. When LDIRECTORD is started, the IPVS routing table will be properly configured.

??? 2) It can be convenient to be started / stopped by HeartBeat.

转载请注明原文地址:https://www.9cbs.com/read-124913.html

New Post(0)