I've resurrected the clonezilla project using the old VM and doc. I've mounted the old BTOPS image and am imaging a client right now to make sure everything still works. I read about proxydhcp last night. Let's use that so we don't have to run our own DHCP server. The plan was to get it working with a USB key but with proxydhcp I don't think we need to do that. Let's jump right to PXE boot.
So the test imaging went perfectly. Now to knock out these:
Image a client to make sure everything still worksPut back in the menu so admins can select the image they wantget proxydhcp workingremove dhcpd and install dnsmasqconfigure dnsmasq for proxydhcptest using an existing DHCP server
Run dcs to try to setup a menu to select images. Total fail. Now the client
doesn't even boot all the way. Running drblpush -i to try to recover. Ok the
recovery went well. Now I'm trying to figure out how to add the menu. Trying
to use the Clonezilla Live option in
DRBL.
Doing a test clone to see where I left off. It's still using the old Debian
image that auto clones. Rerunning drblpush -i. So now clonezilla live is
booting but I don't get a menu, just a unix login prompt.
Looking for clonezilla alternatives. Found Mondo Rescue.
Nope... looks like clonezilla is the magic. Change ocs_live_run in
/tftpboot/nbi_image/pxelinux.cfg/default to ocs-live-general. Can't get the
Clonezilla Live UI to start. The boot is failing at the end with an error about
sudo and mkpasswd. I'm googling trying to fix it now. Let's try upgrading the
whole image to latest. Stopped the upgrade. I added live-config to the boot
params and that worked! Now let's figure out how to preseed the config options.
I think we have to tweak the pxeloader boot parameters to set the menu
the way we want. Setting ocs_live_run to ocs-sr... with ask_user where
we should pause and ask the user seems to do the trick. Now let's get dhcpproxy
working.
Googling dhcpproxy drbl to see if I can get it working. Following these steps: http://sourceforge.net/p/drbl/discussion/Open_discussion/thread/fe0626d0. I had to not use the IPAPPEND 1 option because it messes up the boot. I removed isc-dhcp-server and tftpd-hpa since dnsmasq provides tftp now. Put wildcard host match in /etc/exports so that the exported filesystems would be available to all clients.
So the client boots but can't get an IP address. Played around with IPAPPEND [1|2] which appends stuff to the boot params but still nothing. udhcpc in the initramfs gets an IP address but doesn't set the IP on eth0. Not sure what's going on.
Time to debug proxy DHCP and debian-live IP address stuff. It looks like we can toggle between save and restore using the boot menu(cool)! Here's a guy with the same issue.
I modified Clonezilla-live-initrd.img like this:
diff -Naur Clonezilla-live/lib/live/boot/9990-networking.sh Clonezilla-live_new/lib/live/boot/9990-networking.sh
--- Clonezilla-live/lib/live/boot/9990-networking.sh 2014-08-19 12:23:06.000000000 -0500
+++ Clonezilla-live_new/lib/live/boot/9990-networking.sh 2014-08-19 12:20:52.000000000 -0500
@@ -109,7 +109,11 @@
done
else
for interface in ${DEVICE}; do
- ipconfig -t "$ETHDEV_TIMEOUT" ${interface} | tee /netboot-${interface}.config
+ if [ -n "${STATICIP}" ]; then
+ ipconfig -t "$ETHDEV_TIMEOUT" ${interface} ip="${STATICIP}" | tee /netboot-${interface}.config
+ else
+ ipconfig -t "$ETHDEV_TIMEOUT" ${interface} | tee /netboot-${interface}.config
+ fi
# squeeze
[ -e /tmp/net-${interface}.conf ] && . /tmp/net-${interface}.conf
I setup the save menu and fixed the timeout on boot so now you can save and restore images. Next I should get the IP for pxelinux.cfg/default to change when eth0 gets a new dhcp address. Also auto reboot at the end.
Exporting the virtualbox image as an appliance.
John asked me to set the boot screen timeout to nothing so you have to choose.
John asked me to have flamethrower(multicast) cloning working be the end of next week. I found a really good article on clonezilla multicast restore. It seems pretty simple. The first thing is snapshotting our current clonezilla server image. Next I need to find a test client. I'm using one of the Thinkpad T61's on the desk here. John asked me to save the current image and use that for my restore test.
I don't have a physical client to test so I need to setup a test client
VM. In order to get cloning working between the clonzilla server vm and a
test client vm, I had to set bridge_fd 2 in /etc/network/interfaces.
0 was "out of bounds". Now I can PXE boot a test client VM from the
clonezilla server.
Now we need to figure out this multicast stuff. Running drblsrv -i to
hopefully configure multicast. I think dcs is what we need to run. So dcs
worked. That's definitely the command to run to set up multicast. However it
blows away our custom menu options. So the question is: how do we maintain our
custom menu options while allowing for multicast? Here's the multicast menu
option that got created:
label Clonezilla-live
MENU DEFAULT
# MENU HIDE
MENU LABEL Clonezilla: multicast restore 2015-07-23-lenovo-T400 to disk sda
# MENU PASSWD
KERNEL Clonezilla-live-vmlinuz
APPEND initrd=Clonezilla-live-initrd.img boot=live union=aufs noswap noprompt nolocales netboot=nfs nfsroot=192.168.1.81:/tftpboot/node_root/clonezilla-live/ ocs_server="192.168.1.81" keyboard-layouts="NONE" locales="en_US.UTF-8" ocs_daemonon="ssh" ocs_prerun="mount -t nfs 192.168.1.81:/home/partimag /home/partimag/" live-config ocs_live_keymap="NONE" ocs_lang="en_US.UTF-8" ocs_live_run="ocs-sr -l en_US.UTF-8 -g auto -e1 auto -e2 -r --clone-hidden-data -p reboot --mcast-port 2232 multicast_restoredisk 2015-07-23-lenovo-T400 sda"
IPAPPEND 1
TEXT HELP
Clonezilla Live 2.2.3-25-amd64 runs on RAM
ENDTEXT
It looks like dcs still uses 192.168.8.1 for the ipaddress of the server in
/tftpboot/nbi_img/pxelinux.cfg/default. So if you just go in and edit any
192.168.8.1 address to say 192.168.1.81 instead, it'll work.
John has been having some issues with the server in save/restore unicast mode.
Let's take a look. It turns out that we didn't do anything with the -q ntfsclone option yet so that's not in play. I did a test restore of an image
to one of the Thinkpads and it worked well. I got around 1.35 GiB/min.
I'll ask John to see what he wants to do next. John wants to test
unicast restores on 3 laptops simultaneously.
New idea: Use one clonezilla boot server VM for each network and put the images on a separate multi-homed NFS server VM. Well, it looks like you can have a multi-homed drbl server. So maybe we should just do that.
Anyway, the test unicast went fine. So now I'm trying to figure out how to change the dcs IP from 192.168.8.1 to 192.168.1.81. So I ran this command to find files that might need to be changed:
root@clonezilla:~# grep -r 192.168.8.1 /tftpboot/ /etc | grep -v 192.168.8.10 | cut -f1 -d: | sort | uniq | grep -v drblsave | grep -v bak | grep -v .orig
/etc/dhcp/dhcpd.conf
/tftpboot/node_root/etc/dhcp/dhcpd.conf
/tftpboot/node_root/etc/diskless-image/config
/tftpboot/node_root/etc/hosts
/tftpboot/node_root/etc/network/interfaces
I didn't edit the dhcpd.conf files since dhcpd is turned off. So let's try
running dcs and see what happens. So that seems to have worked. Although I
edited the files above I believe the file that mattered is
/tftpboot/node_root/etc/diskless-image/config.
So to summarize here is how to set up a multicast restore:
- On the clonezilla server VM, run
dcs All Select all the clientsclonezilla-start Start_clonezilla_modeBeginner Beginner mode: Accept the default optionsrestore-disk Restore an image onto the compute-p reboot Reboot client when the clone finishes- Choose the image to restore
sda disk(sd)_disk(a)multicast multicast restoreclients-to-wait Set the number of clients to clone- Set the number of clients you're restoring.
- Press enter through the warnings
- PXE all your clients and select the multicast option
- Once all the clients are at the "wait" screen, cloning will begin.
So next is the ntfsclone
option.
Let's add that. I've add the -q option to both the save and restore
boot menu items in /tftpboot/nbi_img/pxelinux.cfg/default. I'm not
sure if we have to add it to both but I don't think it can hurt.
When trying to save the image of our load balancing router, so I can
use that client for the nextgen project, clonezilla failed to mount NFS.
For some reason the PXE booted client NFS mount fails. I removed IPAPPEND
from /tftpboot/nbi_img/pxelinux.cfg/default and it mounts fine now.
That's probably because there's multiple ethernet adapters in this client.
Well perhaps our custom initrd got stomped and that's why IPAPPEND doesn't
work any longer. I went into ~/clonezilla_live_initrd_edit and ran:
export UNZIP=1
rm -rf Clonezilla-live
make Clonezilla-live
diff -Naur Clonezilla-live/lib/live/boot/9990-networking.sh Clonezilla-live_new/lib/live/boot/9990-networking.sh
After this clone completes we should try putting our custom initrd back in place and PXE booting the client to see if NFS mounts properly then. Also it looks like we're unzipping img.orig while the one that gets used during boot is img. We need to look into that.
So I'm restoring 2016-05-11-DomainClient40gb to the HP tower on the wire rack. John said that he's seen errors with that restore. Ok, so that restore went off without a hitch, booted fine and I was able to login using the standard admin password.
Ok so John thinks the issue may be with the network save. So let's
- verify the currently installed image on the HP Tower in the rack
- save that image via the network
- restore that same image
So the current image on the client is good. It boots into Windows and I'm able to login. Now let's save the image via the network. Ok so the image save is going. It's definitely using ntfsclone which will take a bit longer.