[j-nsp] Routing stopped due to failed hard disk?

[Richard A Steenbergen]

On Sat, May 07, 2005 at 09:51:29AM +0200, Blaz Zupan wrote:
> >First off, you may want to get console on the box and check the logs to
> >figure out what actually happened. Assuming the HD died and really did
> >take down the system, you should be able to boot the device into single
> >user mode (boot -s at the boot loader, just like you would do to recover
> >passwords) and remove the mirror-flash-on-disk command.
> 
> There are no logs - the device on which the logs were located is dead (the 
> hard disk). After a reboot the box came back, but the hard disk does not 
> appear in "show system boot-messages" - only ad0 (the flash) is recognized, 
> ad1 is nowhere to be seen. The mirror-flash-on-disk has been removed and 
> the whole RE will be replaced (hopefully soon). I have also removed 
> mirror-flash-on-disk from all other Juniper boxes, seems like it is a 
> really bad idea.

Well, before you reboot it, you can check dmesg.

The device disappearing is a safety mechanism, to prevent precisely what 
you saw happen. When there really is a problem with a device, it will 
disappear from the device list so that it doesn't keep taking down the 
rest of the system.

You can check/change this from sysctl (or from bios when the device boots 
up if you hit Del fast and often enough). A normal system (well a normal 
old M series, not sure if they changed any names on the Mxi's) should 
have:

machdep.bootdevs: pcmcia-flash,compact-flash,disk,lan

But yeah I've seen too many other much worse problems caused by 
mirror-flash-on-disk, and ccd in general, to trust it. You can do manual 
snapshot updates to keep your alternate media in sync with "request system 
snapshot". Personally I'd be just as happy with a config option to save 
configs to altconfig at commit time, and manual snapshots for the base 
system after you upgrade and confirm that a new version of code works for 
you.

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)