Stopping the CDR service

• The CDR Data Servers
• The CDR Central Server
• CDR Setup
• Starting the CDR service
• Stopping the CDR service
• GRID Proxies
• CDR Tools

To ensure that the file copy processes are handled as "atomic" activities, thus minimizing the probability of getting only part of a file copied to its destination, a soft stop mechanism is in place. The CDR manager can ask each instance of PadmeCDR to stop as soon as it is in a safe state by creating the run/stop_cdr_<SRC>_<DST> files. E.g. touch run/stop_cdr_LNF_CNAF will tell the instance which copies file from LNF to CNAF to stop. Each PadmeCDR instance checks for the existence of its corresponding stop_cdr file every time it is in a safe state: if the file is found, PadmeCDR will delete it and gracefully exit. Note that, depending on the current copy state and on the bandwidth of the network connections, several seconds or even minutes can pass before PadmeCDR reacts to the stop_cdr file.

Of course the good old kill of the PadmeCDR instances still works, but this should only be used as the last resort, i.e. when the "soft stop" mechanism does not work or a gfal command gets stuck. If a PadmeCDR process is kill-ed, the CDR manager should verify from the log file if the last file was completely copied and manually remove any partially copied file from the destination system.

Analogously to PadmeCDR, the CDRMonitor service can be stopped with the command touch run/stop_cdr_Monitor.