openstack
/
fuel-astute
Code Issues Proposed changes

Erase only boot code in MBR during node delete 

Zeroing first 1Mb of a block device will almost always
lead to serios system error, if there is a filesystem
on this block device and this filesystem is mounted.

This code had 2 points: to prevent system from booting
from hard drive after reboot and to prevent anaconda
(or other OS installer) to automatically recognize
disk structure. But since we only use image-based provision
this is not the case anymore.

It's enough to clear first 446 bytes of MBR where the first
stage of a bootloader resides. So we will also be able to
check nodes disks and filesystems after deleting it from
cluster for troubleshooting for example, because partition
scheme will not be touched.

https://wiki.archlinux.org/index.php/Master_Boot_Record

Change-Id: I60f2b250d6d5997b4b471c27ed00d057ae73c30a
Fixes-bug: #1538587
This commit is contained in:
Dmitry Guryanov 2016-03-25 16:19:29 +03:00
parent f7e6967063
commit e770d4ec7d
1 changed files with 4 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
mcagents/erase_node.rb
View File

@ -112,19 +112,10 @@ module MCollective
File.open('/proc/sys/kernel/panic','w') {|file| file.write("10\n")}
begin
# Delete data disks first, then OS drive to address bug #1437511
(get_devices(type='data') + get_devices(type='root')).each do |dev|
debug_msg("erasing #{dev[:name]}")
# try an make sure any ext2/3/4 paritions won't cause a panic
# when the file system gets erased
suppress_fs_panic(dev[:name])
# clear out the partitions of the device
erase_partitions(dev[:name])
# attempt to clear our the file system data on the device partitions
# to ensure that we aren't left with extra filesystem data when we
# reparition the devices next time.
erase_data(dev[:name])
erase_data(dev[:name], 1, dev[:size], '512')
get_devices(type='all').each do |dev|
debug_msg("erasing bootstrap code area in MBR of #{dev[:name]}")
# clear out the boot code in MBR
system("dd if=/dev/zero of=#{dev[:name]} bs=446 count=1 oflag=direct")
end
reply[:erased] = true
@ -149,30 +140,6 @@ module MCollective
Process.detach(pid)
end
def suppress_fs_panic(dev)
Dir["/dev/#{dev}{p,}[0-9]*"].each do |part|
# if the partition has ext2/3/4, make sure error action is set to
# continue and not panic so that we can erase it correctly.
if system("blkid -p -n ext2,ext3,ext4 #{part}") == true && system("findmnt -ln -S #{part}") == true
system("mount -o remount,ro,errors=continue #{part}")
end
end
end
def erase_partitions(dev)
Dir["/dev/#{dev}{p,}[0-9]*"].each do |part|
system("dd if=/dev/zero of=#{part} bs=1M count=10 oflag=direct")
end
end
def erase_data(dev, length=1, offset=0, bs='1M')
cmd = "dd if=/dev/zero of=/dev/#{dev} bs=#{bs} count=#{length} seek=#{offset} oflag=direct"
status = system(cmd)
debug_msg("Run device erasing command '#{cmd}' returned '#{status}'")
status
end
# Prevent discover by agent while node rebooting
def prevent_discover
lock_path = AGENT_NODISCOVER_FILE