Automatic Restart of Services Using Watcher

Calendar Server and Messaging Server now use the same stop and start mechanism. The start-cal and stop-cal commands are wrappers for a new internal service, csservice, which was introduced as part of the Watcher implementation. This service starts the Watcher, and then starts all other processes. The csservice program is aware of any dependencies the other services have, and in which sequence the services should be started.

Each registered service (process) opens a connection to the Watcher. If a process dies without properly disconnecting, the Watcher automatically restarts it. If the process dies twice in a defined interval, Watcher does not restart it. This timeout interval is configurable.

Additional Watcher information:

• Calendar Server Services Monitored by Watcher

• Configuring Watcher

• Watcher Logging

• Automatic Restart in High Availability Deployments

• Wrapper Scripts for csservice

Calendar Server Services Monitored by Watcher

The Watcher monitors all of the services registered with it. For Calendar Server, the registered processes are: cshttpd, csadmind, csdwpd, dsnotifyd.

If csstored is enabled, that is, if the configuration parameter local.store.enable is set to "y", then csstored is also registered with the Watcher. When it is enabled, csstored must be successfully started before each service that accesses the store can be started. If it stops, then the dependent processes must be stopped an restarted also.

Configuring Watcher

Watcher is enabled by default. To manage the Watcher process, new parameters were added to the ics.conf file:

• local.watcher.enable = "y": the start program (csservice) attempts to start the Watcher before any other services. If this parameter is set to "n", then the Watcher program is disabled.

• service.autorestart = "y": the Watcher automatically restarts stopped services. If set to "n", Watcher does not restart stopped services. If this parameter is set to "n" , Watcher still monitors the services and sends failure or non-response error messages to the console and the cal_svr_base/data/log file.

• local.autorestart.timeout = "600": the default time within which a second server failure triggers Watcher to stop trying to do a restart.

• local.watcher.port: the default port is "49994"; however, if you have Messaging Server, it will also be listening on this port and will be in conflict with Calendar Server. To avoid possible conflict, it is safer to choose a different port for Watcher to listen on.

Watcher Logging

Watcher writes to two logs:

• cal_svr_base/data/log: watcher sends failure notices and non-response error messages to the console. These messages are also written to this log.

• cal_svr_base/data/log/watcher: watcher records all server stops and starts in this log file.

Automatic Restart in High Availability Deployments

If a server fails twice within the timeout period, the system stops trying to restart the server. In an HA system, Calendar Server is shutdown and a failover to the other system occurs.

Wrapper Scripts for csservice

The public interfaces to csservice are start-cal and stop-cal. This section shows the usage for each of these wrapper scripts and contains tables with explanations of their options and a list of components to be started or stopped.

start-cal Wrapper Script

The start-cal usage is as follows:

./start-cal [options...] [components...]

The following is the list of options:

-? or --help

Display this help list.

-d

Enable debugging mode.

-l

List active services.

-L

List enabled services.

-A

List all services.

This following is the list of components:

If no components are listed, start-cal starts all enabled services.

stop-cal Wrapper Script

The stop-cal usage is as follows:

./stop-cal [options...] [components...]

The following is the list of options:

-? or --help

Display this help list.

-d

Enable debugging mode.

-f

Force stop using SIGKILL. (This works only with UNIX® platforms.)

This following is the list of components:

If no components are listed, stop-cal stops all enabled services.