Failover conditions: What counts as Failed?
Wavestore systems support Failover Groups, where one or more Standby Servers can take over from the main servers if the main servers fail.
But what do we mean by "Fail" in the context of Failover?
Failure Conditions
A number of conditions are considered to be Failed for Failover purposes:
- The server loses power or stops working and vanishes from the network
- The server reboots or restarts
- A network interface or bond group goes down. This could be because a cable has been disconnected. Note: if a single link of a network bond group goes down this does not count as failed, all links of the network bond group need to be down. See configuration options below.
- A disk or disk array fails, so no data can be written to the disk or disk array. Note: if a single disk of a disk array goes down and redundant disks mean it continues to work then this does not count as failed, only if the array is no longer functional will it count as failed. See configuration options below.
- A Failover event action is performed. This event action allows the user to configure an event which can trigger the server to be treated as failed.
- A force-failover command has been executed. This allows the user to force failover for maintenance purposes.
This list is correct for version 6.34 and later.
Recovery from Failover
In the case of a reboot or restart, or network or disk failure, the server will be counted as working again once the reboot/restart/failure has been resolved. In the case of a Failover event action or force-failover command, it will be necessary to restart the server to recover from the condition.
Configuration Options
Network Interface and Disk Array failure condition checks can be disabled by configuration options (Netcheck=n and Diskcheck=n). These need to be manually applied into the configuration file if needed. Disabling these checks is not recommended.