Opened 12 years ago

Closed 12 years ago

Last modified 12 years ago

#499 closed defect (fixed)

'exit' to reboot doesn't work correctly

Reported by: Bruno Cornec Owned by: Bruno Cornec
Priority: normal Milestone: 3.0.0
Component: mindi Version: 2.2.9.7
Severity: normal Keywords:
Cc:

Description

Reported to me:

I made a mondoarchive to a DVD. (Proliant DL380 G5). Booting from that DVD goes fine, I can launch the "compare" that completes fine but when typing 'exit' to reboot, it goes wrong: this causes a Kernel panic...

Idem if I boot DVD and select "expert" mode, selecting 'exit' from cli gives a Kernel panic as well...

Is there another way to 'nicely' reboot the box from the normal harddisk after such operation other than power-cycling the box or selecting "Reset" from the ilo-power button?

I've also seen that with test on RHEL 6 in a QEMU VM.

Change History (5)

comment:1 by Bruno Cornec, 12 years ago

Status: newassigned

This is clearly a bug that I've also seen recently with the latest version.

Maybe the reboot/halt command could be called instead ? iLO reset is of course a good solution, or power cycle from it or directly of course, but that should be fixed anyway.

comment:2 by Bruno Cornec, 12 years ago

Another feedback:

  • exit gives a kernel panix
  • reboot does nothing
  • shutdown -ry now restarts init processs; Welcome to init (from mindi 2.0.7.8-r2840) and starts the mondorescue shell.
  • halt does indeed halt the systeem but now need to powercycle the vmware guest or server.

comment:3 by Bruno Cornec, 12 years ago

Component: mondomindi

this has been identified and reproduced. Busybox in this new version has changed its behaviour when init exits. Working on improving this script to support correctly busybox 1.18

comment:4 by Bruno Cornec, 12 years ago

Resolution: fixed
Status: assignedclosed

Rev [2888] should fix this problem for mindi-busybox 1.18.5. Packages with this fix are available for test at ftp://ftp.mondorescue.org/test/ to confirm. However in the VM, as we use sysrq it doesn't kill the VM, but I'm not usre this is a big issue here. May work on a diffrent approach (using a real init and moving the current one into rcS) later on.

A work-around is to use mindi-busybox 1.7.3 with latest version of mondo which doesn't expose this issue.

in reply to:  4 comment:5 by Bruno Cornec, 12 years ago

I had to change completely te way init is managed with the new mindi-busybox (using a real init from busybox and moving our init script to etc/init.d/rcS). It's now available with rev [2893].

The workaround with mindi-busybox 1.7.3 doesn't seem to work for eveybody.

Packages with this fix are available for test at ftp://ftp.mondorescue.org/test/ to confirm (date 20111018).

Note: See TracTickets for help on using tickets.