Fabric OS Troubleshooting and Diagnostics Guide v7.0.0 (53-1002150-02, June 2011)

14 Fabric OS Troubleshooting and Diagnostics Guide
53-1002150-02
Switch boot
2
Rolling Reboot Detection
A rolling reboot occurs when a switch or enterprise-class platform has continuously experienced
unexpected reboots. This behavior is continuous until the rolling reboot is detected by the system.
Once the Rolling Reboot Detection (RRD) occurs, the switch is put into a stable state so that a
minimal supportSave can be collected and sent to your service support provider for analysis. Not
every reboot activates the Rolling Reboot Detection feature.
ATTENTION
If a rolling reboot is caused by a panic inside Linux kernel, then the RRD feature is not activated.
Reboot classification
There are two types of reboots that occur on a switch and enterprise-class platform, expected and
unexpected. Expected reboots occur when the reboots are initialized by commands, these types of
reboots are ignored by the Rolling Reboot Detection (RRD) feature. They include the following:
• reboot
• haFailover
• fastBoot
• firmwareDownload
The RRD feature is activated and halts rebooting when an unexpected reboot reason is shown
continuously in the reboot history within a certain period of time. The period of time is switch
dependent. The following are considered unexpected reboots:
• Reset
A reset reboot may be caused by one of the following:
- Power-cycle of the switch or CP.
- Linux reboot command.
- Hardware watchdog timeout.
- Heartbeat loss-related reboot.
• Software Fault:Kernel Panic
- If the system detects an internal fatal error from which it cannot safely recover, it outputs
an error message to the console, dumps a stack trace for debugging, and then performs
an automatic reboot.
- After a kernel panic, the system may not have enough time to write the reboot reason
causing the reboot reason to be empty. This is treated as an Unknown/reset case.
• Software fault
- Software Fault:Software Watchdog
- Software Fault:ASSERT
• Software recovery failure