|
|
Andrew Kaplan
|
|
Oct 21, 2009 18:26:12 GMT
|
|
Hi there --
I am trying to do a restore of an image to a workstation that recently had its hard drives replaced due to failure. The replacement drives are SAS types, and did not have any filesystems formatted on them when they were installed on the workstation. The .iso files total fourteen in number and each are about 650 megabytes. When I insert the first CD into the workstation's DVD/RW drive, the boot sequence commences but then hangs at the point where the message:
Freeing unused kernel memory: 236k freed
appears on-screen.
I let the system stay in that status for about five minutes before having to reboot the workstation. When an operating system CD is installed into the workstation, it is able to detect the harddrives without any difficulty.
What is going on here, and what can I do to correct it? Thanks.
|
|
|
Note: If you are the author of this question and wish to assign points to any of the answers, please login first.For more information on assigning points ,click
here
|
|
|
Sort Answers By:
Date or Points
|
|
Steven E. Protter
|
|
Oct 21, 2009 20:15:40 GMT
10 pts
|
|
Shalom,
mondorescue is not going to be able to detect the disk unless its the same type of disk that the system had when it was imaged.
The OS recognizes the hardware, mondorescue does not.
What you can do is:
Base OS install.
mount -o loop filename <mount point>
From the iso files you should be able to extract restore information.
SEP |
|
|
Andrew Kaplan
|
|
Oct 21, 2009 20:24:07 GMT
N/A: Question Author
|
|
Hi there --
Thanks for your reply. Extracting the files is one thing, but if I want to do the actual installation of all fourteen .iso files onto the system. It seems I would I need to take to accomplish the goal of restoring the workstation to its original configuration. After extracting the files, how can I do the restoration? |
|
Court Campbell
|
|
Oct 22, 2009 14:31:00 GMT
10 pts
|
|
for the future you will want to edit /usr/sbin/mindi by editing the FORCE_MODS line to be
FORCE_MODS="diskdumplib diskdump mptbase mptscsi mptspi mptsas mptscsih ata_piix crc32 mii
pcnet32 sunrpc nfs nfs_acl lockd loop"
More than likely you do not have mptsas built into your initrd for mondo. |
|
|
Court Campbell
|
|
Oct 22, 2009 14:40:12 GMT
10 pts
|
|
|
Do you have another machine that has the same distro and kernel version running, etc? One thing you could try, if you have a nearly identical setup, is to run mindi on that host to create an initrd with the necessary mptsas module. Then you could extract the 1st iso, add the new initrd, and recreate the iso. I know that is involved, but if you go the time... |
|
|
Court Campbell
|
|
Oct 22, 2009 15:11:21 GMT
10 pts
|
|
|
I've been trying to think of ways to attack this. I haven't had this specific issue because I always changed my mindi script to add more modules. I use it a alot for creating clones to virtual machines. Anyway, another thought I had was to boot from a live cd. Then you could create the filesystem on the sas drive. Next install mondorescue in the live environment. Then run mondorestore with the -i switch and point it to the iso. You should then be able to edit the disk portion to point to the new drive. You might need a rescue cd afterwards to fix grub. But I believe that would work. |
|
|
Andrew Kaplan
|
|
Oct 22, 2009 17:09:33 GMT
N/A: Question Author
|
|
Hi there --
I tried creating a mindi boot disk on a similar system with the additions to the FORCE_MODS line that were suggested in an earlier posting. I then created the mindi boot disk, and brought it to the system in question. The same hanging situation that I described in my original posting occurred with the mindi boot disk.
If I try booting with the livecd, am I doing an install of it onto the hard drive or am I simply running it in RAM, and doing the mondorestore from there? |
|
|
Court Campbell
|
|
Oct 22, 2009 20:23:51 GMT
10 pts
|
|
>> When an operating system CD is installed into the workstation, it is able to detect the harddrives without any difficulty.
What OS cd? Also what distro are you running (version number w/ update number)? |
|
|
Andrew Kaplan
|
|
Oct 22, 2009 20:55:09 GMT
N/A: Question Author
|
|
The OS cd is Fedora Core 2. The distro of Mondorescue is version 2.2.8.
I had one idea: boot the system with a Knoppix LiveCD, mount the hard drive devices, and do the install that way.
That does bring up another question: Assuming that Knoppix does not come with mondorescue, how can I install or run the restore CD's? |
|
|
Court Campbell
|
|
Oct 23, 2009 13:21:14 GMT
10 pts
|
|
|
I am going to do a test today to see what it would take to get this to work from a live cd. I do think there will still be an issue of the restored initrd not having the module to see the disk. So you will need a rescue cd. Let me see what I find out. |
|
|
Andrew Kaplan
|
|
Oct 23, 2009 15:19:42 GMT
N/A: Question Author
|
|
Hi there --
Thanks for your reply. I will be waiting in anticipation of your test results. Suffice it to say, your efforts in this regard are greatly appreciated. |
|
|
Court Campbell
|
|
Oct 23, 2009 17:45:02 GMT
10 pts
|
|
Andrew,
I just want to clarify something. The OS the the workstation was running is FC2, correct?
Also, on the other host that you used mindo to create another boot iso, is that also FC2? If so can you logon to that box and see if you can modprobe mptsas? I am wondering if FC2 doesn't have that module. |
|
|
Andrew Kaplan
|
|
Oct 23, 2009 18:18:39 GMT
N/A: Question Author
|
|
Andrew,
I just want to clarify something. The OS the the workstation was running is FC2, correct?
Also, on the other host that you used mindo to create another boot iso, is that also FC2? If so can you logon to that box and see if you can modprobe mptsas? I am wondering if FC2 doesn't have that module.
Hi there --
The OS the workstation was running is indeed the FC2 distro.
Also, I ran the modprobe command on the other host, modprobe -l |grep mpt |more, and that particular one did not have the mptsas.ko module.
I found another workstation that did have the module, and I am running mindi on that system. Once that is done I will create the iso image, and try it on the problem workstation.
I will post the results as soon as I get them. |
|
|
Court Campbell
|
|
Oct 23, 2009 18:26:09 GMT
10 pts
|
|
|
Just don't forget to edit mindi on that host, or ad least run modprbe mptsas before making the image. |
|
|
Court Campbell
|
|
Oct 23, 2009 20:44:14 GMT
10 pts
|
|
All right. I got it to work with a live cd (for the most part). I am attaching a doc with the steps I took. It's not very detailed. I assume that you can follow it. If you have any questoins, just post them and I will try and help.
One issue I had is that I left a line in /etc/fstab that I did not create and restore during the mondorestore process. So I had to use the rescue cd to reboot and edit the file. But that is because I couldn't remember the root password. Other than that it came up fine. But I also wasn't restoring to a different disk type.
Good luck! |
|
|
Andrew Kaplan
|
|
Oct 24, 2009 00:34:39 GMT
N/A: Question Author
|
|
Hi there --
I began going through the procedure you sent, thanks by the way, and I ran into my first issue.
At the fdisk step, I identified the disk in question as /dev/sda. I ran the pvcreate, and vgcreate commands using the suggested syntax without issue.
When I got to the lvcreate command, the syntax that is shown returned an error indicating I needed to include the size of the logical volume, or -L argument in the command. When I ran the command with the -L argument, I specified the size at 271000 megabytes. The error message I got was there was not enough physical extents available.
Did I miss a step here? |
|
|
Court Campbell
|
|
Oct 24, 2009 02:16:08 GMT
10 pts
|
|
Like I said, the doc does not have everystep. You should have used fdisk and created two primary partitions. The first one for /boot and the other for LVM. You also need to change the type of /dev/sda2 to LVM. That is 8e in fdisk. In the past I have had LVM issues if i forgot to change the type. Also when using lvcreate with -L you can use size notations to create then, like so:
lvcreate -n LogVol00 -L +2G VolGroup00
that will create a 2GiB lvol. Also, are you just creating a single large lvol? If so just run vgdisplay VolGroup00. Then look at the free pe line. Then use lvcreate with a -l instead and give it the free pe number from vgdisplay. |
|
|
Andrew Kaplan
|
|
Oct 24, 2009 03:45:14 GMT
N/A: Question Author
|
|
Hi there --
Sorry about my ignorance. I don't often create logical volumes on our Linux systems. Normally that honor is reserved for our HP-UX servers.
I'll pick up where we left off on Monday. Have a good weekend. |
|
|
Andrew Kaplan
|
|
Oct 26, 2009 14:25:55 GMT
N/A: Question Author
|
|
Hi there --
I went through the procedure this morning, and everything appeared to work fine until the reboot, or lack of it, took place.
When I did some further investigating, somehow everything got put on /dev/sdb instead of /dev/sda. I am positive that I typed everything in correctly. How this happened I do not know.
I am going to some additional work on my end, but I wante to see if you had any ideas? Thanks. |
|
|
Court Campbell
|
|
Oct 26, 2009 14:37:57 GMT
10 pts
|
|
I can only assume that you partitioned /dev/sdb, created the lvm structures, etc. and mounted it.
And since you don't have the correct module for your drive added to your initrd from the mondobackup, I would not expect it to boot. You will need to use an fc2 install cd and run linux rescue. |
|
|
Andrew Kaplan
|
|
Oct 26, 2009 20:10:07 GMT
N/A: Question Author
|
|
Hi there --
I tried several more times to recover the system without any success. I have no choice at this point than to rebuild the system from scratch.
After that is done I am going to have to rethink how the backups are done using mondo in order for this incident not to occur again.
Thanks anyway for your help. |
|
|
Court Campbell
|
|
Oct 26, 2009 20:17:24 GMT
10 pts
|
|
Sorry it's not working. I wish I could be there, I really love challenges like this one.
If anything, just do a fresh load of fc2, and them load mondorescue. Then use mondorestore to restore all the users data. That should work fine. My suggestion would be to not use mondo to backup your whole machine. I only use it to backup the main system directories. Then I use normal backups for everything else. |
|
|
Andrew Kaplan
|
|
Oct 26, 2009 20:40:04 GMT
N/A: Question Author
|
|
Hi there --
I posted the question as part of the mondo-devel users group, and one suggestion that was made was to use the following syntax at the boot prompt of the mondo restoration cd:
denymods=fdomain
At the risk of sounding like an idiot, how would I go about determining the name of the kernel to enter at the prompt in conjunction with the above parameter?
Also, why would fdomain be causing the system to hang at freeing unused kernel memory? |
|
|
Court Campbell
|
|
Oct 26, 2009 20:45:41 GMT
10 pts
|
|
You shouldn't need to worry about the kernel. just type this at the boot.
nuke denymods=fdomain
I am not sure if that will work, but it's worth a shot. |
|
|
Andrew Kaplan
|
|
Oct 27, 2009 13:23:30 GMT
N/A: Question Author
|
|
Hi there --
The nuke value at the boot prompt refers to the name of the kernel. If I knew the name of the kernel, I could insert that along with the denymods=fdomain argument.
I have a system that is the same model and operating system of the problem workstation. Is it possible to reference that system to get a clue to the name of the kernel? |
|
|
Court Campbell
|
|
Oct 27, 2009 13:26:43 GMT
10 pts
|
|
|
You've lost me on that one. I have never typed a kernel name when using mondo. I usually use nuke or interactive. Have you tried running nuke denymods=fdomain? |
|
|
Andrew Kaplan
|
|
Oct 27, 2009 13:45:21 GMT
N/A: Question Author
|
|
Hi there --
I did try the nuke denymods=fdomain, and that is when I got confronted with the error indicating that mindi could not find a kernel with the name nuke. |
|
|
Court Campbell
|
|
Oct 27, 2009 15:30:23 GMT
10 pts
|
|
try
vmlinuz denymods=fdomain |
|
|
Andrew Kaplan
|
|
Oct 30, 2009 20:32:59 GMT
N/A: Question Author
|
|
Hi there --
I ended up rebuilding the workstation from scratch, and now comes the fun part: determining what needs to be done in order to prevent this incident from reoccurring.
Enclosed with this posting is a copy of the script that is used to create the .iso images.
The script in its current form creates the backups based on the filesystems of the workstation.
I had several ideas concerning the script, and the backup:
1. Going by your example, modify the mindi file, specifically the FORCE_MODS line, on all workstations to include the arguments:
FORCE_MODS="diskdumplib diskdump mptbase mptscsi mptspi mptsas mptscsih ata_piix crc32 mii pcnet32 sunrpc nfs nfs_acl lockd loop"
2. Adding a section to the script that will run mindi, and place the created boot files at the same NFS location where the iso images are located.
Besides or instead of the above, is there anything else that I should consider putting into the script? |
|
|
Court Campbell
|
|
Oct 30, 2009 21:49:37 GMT
10 pts
|
|
Andrew,
Send me your e-mail address. Send it to thouaeti_at_guerrillamailblock_dot_com.
That's a temporary address. It might be easier than using this thread.
Later,
Court |
|
|
Andrew Kaplan
|
|
Nov 2, 2009 01:45:30 GMT
N/A: Question Author
|
|
Hi Court --
I tried the address you provided, and was not able to connect to it. Here is my address: ahkaplan@partners.org
Thanks. |
|
|
Andrew Kaplan
|
|
Nov 3, 2009 00:39:13 GMT
N/A: Question Author
|
|
Hi Court --
Did you receive my e-mail address? I had not heard from you, so I am just verifying you had received it. Thanks. |
|
|
Steven E. Protter
|
|
Nov 3, 2009 05:49:44 GMT
10 pts
|
|
Court has been busting his hump on this.
Maybe hand him some points.
I'm reading this but want to build on my original suggestion.
Do a base OS install on the system with fresh media. You may not be able to use SAS drives because FC2 does not support that.
Then mount -o loop name.iso /mountpoint
And recover the necessary files and data that way.
SEP |
|
|
Court Campbell
|
|
Nov 3, 2009 06:10:30 GMT
10 pts
|
|
Andrew,
Got your address. just been busy. I am having a data protector issue that I need get fixed. |
|
|
Andrew Kaplan
|
|
Nov 3, 2009 12:34:36 GMT
N/A: Question Author
|
|
Hi Court --
I work with Data Protector myself. Maybe I can lend a hand. What's going on? |
|
Printable version
