Changes of the server states (SS, HS, SBS, CS) are also characterized by changing its states into temporal states, which are specified by RD parameters given in seconds. The list of temporal states is shown in the following table:
State | Meaning |
---|---|
iNone | Stable state |
iElection | Election |
iWaitingHot | Waiting for HS |
iWaitingReadyHot | Waiting for ready HS |
iStartingKernelToSBS | Starting the server to SBS state |
iStartingKernelToHOT | Starting the server to HS state |
iHotOrSBSToSBS_WaitForHot | Waiting for HS after required change |
iHotOrSBSToSBS_WaitAnsConn | Waiting for confirmation of SBS log on to HS |
The basic assumption for the correct setting is to know the time (T_START), that is required for starting the server on particular platform. The server start time is the time from starting the server to the moment when the server is able to connect to a client. It depends mainly on the configuration database size, which is read by the server. You must realize, that the time increases along with extending the application.
Stable state. The server in this state has already read the configuration database. If the server is SBS, the configuration database is synchronized with the HS configuration database. Next configuration changes are replied in real time.
The state, when a HS is not selected. This state is the time, when the servers, which are able to change their states into the HS state, swap own RDS parameters (using MULTICAST) in order to choose a new HS. The state is terminated by:
Recommended period of this state is 5 to 10 seconds (10 seconds for OpenVMS platforms).
RD_TIMEOUT_iElection = 7
The time period, when the server is waits for the HS registration. Server is in this state when:
Recommended minimum period of the state is
RD_TIMEOUT_iWaitingHot = RD_TIMEOUT_iElection * 1.25.
It is 25 % longer than the iElection state.
The time when the server waits for the ready HS. Server is in this state, when it is known which server is to be the HS, but the HS is not completely ready for its functioning (cannot connect a client), because is reading the configuration database. The time period of the state is at least the time T_START. The state is terminated, when the HS notify the other server that it is completely ready. If the server is in the state after unexpected failure of the HS, the server automatically changes its status to the CS status (it will be restarted).
Recommended minimum period is T_START*1.5.
RD_TIMEOUT_iWaitingHot = T_START*1.5
The time when the server:
The configuration database synchronization time strongly depends on the size of the databases and the differences between them. If the server is stopped for a short time, the configuration are almost identical or minimal and the synchronization time is reduced to reading the databases.
Required minimum period of the state is T_START*2.5.
RD_TIMEOUT_iStartingKernelToSBS = T_START * 2.5
The server is in the state after:
The server reads the configuration database in the state. The time period of the state is at least T_START.
Recommended minimum period of the state is T_START*1.5.
RD_TIMEOUT_iStartingKernelToHot = T_START * 1.5
Operator switches one server manually to the HS state. All servers, except the switched server, are to be in the state. The state is the time when the expectant SBS servers are waiting for confirmation of the HS state. Practically, the change takes a short time and the period should be set to the value equal to RD_TIMEOUT_iWaitingHot.
If a server is still in the SBS state and is waiting for confirmation or establishing the connection to the HS after manual switching the HS, then we can say the server is in the state iHotOrSBSToSBS_WaitAnsConn. Practically, the change takes a short time and the period should be set to the value equal to RD_TIMEOUT_iWaitingHot.