For better efficiency and more precision in monitoring the state of the servers, the Watchdog is recommended as a separate component of any HotStandby configuration.
If only two machines are available, making it impossible to run the Watchdog program in a separate machine, run the Watchdog on the same machine where the Secondary server resides and set the parameter AutoPrimaryAlone to no in the configuration file (solid.ini) of both the Primary and Secondary server. Note that setting this parameter to no is extremely important, as it prevents the potential error of having two Primary servers.
CAUTION If both servers are in a state that allows writing (PRIMARY ALONE or STANDALONE), and if the databases of both servers are independently updated, then it will not be possible to resynchronize the two databases. Make sure that the Watchdog does not allow both servers to be put in the PRIMARY ALONE or STANDALONE state at the same time. See Network partitions and dual primaries.
If the Primary server does fail, then the Watchdog is able to switch the Secondary to become the new Primary.
There are some disadvantages to putting the Watchdog in the same machine as the Secondary. The disadvantages include:
▪If only the communication link between the Watchdog and the Primary is down, this configuration may result in a false switchover between the Primary and the Secondary.
▪The communication link becomes a “single point of failure”, that is, a single failure that may disable the entire system. (In most HotStandby configurations, the entire system is not disabled unless there are at least two failures.)
▪If there is a network failure and the Secondary machine cannot communicate with the Primary machine, the users and applications are still able to access the Primary server and theoretically could continue operating with the Primary server. However, the Primary server stops accepting transactions because the watchdog cannot notify the Primary server to continue operating, for example by switching to PRIMARY ALONE state.