White Papers

With a system reload, the system must read and apply the entire startup-config file, which might take some time if the
startup-config is large. Restarting a process saves time because only a portion of the configuration related to the crashed
process is read and reapplied.
For a dual-RPMs system, restarting a process also precludes launching the failover process on the primary and standby RPMs.
Recovery is attempted first locally on the primary RPM, which involves less CPU overhead, increasing the systems availability
for other activities.
However, in both single and dual-RPM systems, even when you configure process restart, the coredump portion of failover is
still executed.
The processes that you can restart fall under three categories:
● Interface-related processes — TACACS+, RADIUS, CLI, and SSH, and so on.
● Protocol tasks — OSPF, RIP, and ACL, and so on. Process restart is not currently available for protocol tasks; the failover
procedure is executed immediately after software exception.
● Line card processes — IPC, Event Log Agent, Line Card Manager, and so on. Process restart is not currently available for
line card processes; the failover procedure is executed immediately after software exception.
Enabling Process Restartability
The restart time varies by process.
In general, interface-related processes are hitless and can be restarted in seconds; if a restart is successful, traffic is not
interrupted. Protocol tasks and line card processes are not hitless and take longer to restart. You can select which process may
attempt to restart and the number of consecutive restart attempts before failover, but by default, every process fails over.
● Enable process restartability for a process or task.
CONFIGURATION mode
process restartable [process] [try number] [timestamp hours]
● Display the processes and tasks configured for restart.
EXEC Privilege
When a process restarts, FTOS displays a message similar to the following message.
[9/18 23:22:21] TME-(tme): Starting to restart the failed process tacplus
[9/18 23:22:41] TME-(tme): Finishing restarting the failed process tacplus
You can specify the timestamp in hour(s) so that if the number of attempts to restart exceeds the maximum allowed within this
timestamp, Restart mode is changed into Failover mode from that moment forward. This means that the next time the crashed
process does NOT restart but failover to the standby RPM if it is on a dual RPM environment and rebooted if it is on a single
RPM.
352
High Availability (HA)