[j-nsp] Junos power-off not graceful

[Phil Rosenthal]

> On Jul 29, 2015, at 5:53 AM, Mark Tinka <mark.tinka at seacom.mu> wrote:
> 
> 
> We once experienced a complete power outage to some MX480 devices that
> caused MPC failure. Those had to be replaced.
> 

I would suspect that this was caused by a power fluctuation just before the power outage -- a surge or brownout. 

> Other items at risk could be hard drives inside the RE.
> 
> It's hard to say - in some cases, you may be lucky, you may not be.
> Hence the "graceful" to increase your chances of a good outcome.

During early testing before adding MX960 to our network, we did have a "pull the plug" test on a chassis, and had no issues of any hardware failure.

One thing you will absolutely encounter, however, is a longer boot time on the following boot as a fsck must be completed on any filesystem not properly unmounted.  There is of course a risk that if the router happens to be making a large write at the exact moment of power loss, that the fsck will not be able to automatically complete, and you will need to manually push it through this process, so be prepared with a serial console afterwards.

If this type of power-off scares you, this should point to changes in your disaster procedures.  In the event of a real disaster, you will not have time to manually do a shutdown -- so I would suggest writing software that is triggered by a low charge level from your UPS system and a Generator failing to start, which causes the MX to have a graceful shutdown. In the event of a UPS failure, you will obviously not have any time to make even an automated shutdown, and this type of "pull the plug" test is what will happen.

Best Regards,
-Phil Rosenthal
ISPrime