Error trying to restore GPT partition tables. Exit returned code 4

[Quazz]

@Matthieu-Jacquart 1903 is no problem on my end, though I must highlight I’m confused by your partition layout.

On a standard W10 GPT install you’d expect 4 partitions (minimum), and on a standard MBR install you’d expect 2.

What FOG version you on?

[Matthieu Jacquart]

@Quazz It’s fog 1.5.6.4 with latest kernel (4.19.36 and 4.19.48).
Indeed I’ve got 2 images with 4 partitions (one is hidden 16MB, not showed on disk manager but listed with diskpart) and no problem to deploy these ones, but on 2 other images there’s just 3 partitions (OEM, system and windows) and I’ve got trouble on different configuration…
So it may be related to mbr2gpt command I use to transform legacy to UEFI but I made it for my 4 images and 2 only are ok, I don’t know why and how to fix that…

[Sebastian Roth]

@Matthieu-Jacquart Beside the layout being a bit weird with the three partitions it actually looks very good for being properly re-sizable! The smaller unmovable partitions are in the front (sda1 and sda2) while the big Windows partition is last.

Interesting that you were not able to deploy this. Let me ask you a few questions:

1. What size is the disk you captured the image from?
2. What size is the disk you tried to deploy it to?
3. Can you please schedule another deploy on this machine but schedule as debug this time!? Start the task as usual till you hit the error and get back to the shell. Now run the following command
4. Can you please upload d1.mbr of that exact image to a fileshare and post a link here?

This one you just posted reminds me on a topic that we had some weeks ago. In the end I think it turned out to be a very strange partition start number issue that was only in the d1.mbr file but not in the text-based partition layout files. I still have no idea why something like this would happen: https://forums.fogproject.org/topic/13220/error-trying-to-restore-gpt-partition-deploying-an-image-to-smaller-disk

[Matthieu Jacquart]

@Sebastian-Roth Image was captured on 128GB SSD and failed to deploy on 120GB SSD, but I found an ugly solution that resolved my problem very well : thanks to you and @Quazz I did’nt notice that good layout contains 4 and not 3 partitions, so I just copy all files from a “good” image to a bad image folder, exept windows data partition that I renamed from d1p3.img to d1p4.img and everything works great !!!

Defective image

Fonctionnal image

New image by copy all files from fonctionnal image except d1p4.img from defective image

So for me this problem is resolved, big thanks to Fog team
Matthieu

[Sebastian Roth]

@Matthieu-Jacquart Wow that’s a real nasty hack to just combine the images. If it works for you - great - go ahead!

Image was captured on 128GB SSD and failed to deploy on 120GB SSD

Yeah I guess this is why it fails. Not because your partition layout is not re-sizeable but because some strange thing in the d1.mbr is causing the issue I reckon.

[Matthieu Jacquart]

@Sebastian-Roth Hi Seb, finally my trick resolve issue on some smaller disk (works now on 120GB disk with image created on 128GB disk), but some old computer with 60GB have still same error, can you take a look to d1.mbr file here ?
https://send.firefox.com/download/f8a134b57bdefad8/#lyrswaozWAFfuWTEcrYrNQ
Maybe some hex modification as in this topic is possible ? https://forums.fogproject.org/topic/13220/error-trying-to-restore-gpt-partition-deploying-an-image-to-smaller-disk/7

Thanks
Matthieu

[Sebastian Roth]

@Matthieu-Jacquart Sorry for the long delay but I have not found the time to look into this any ealier. The d1.mbr file you posted shows the following partition table:

label: gpt
label-id: 4C01BDD7-29D4-4B8C-99D8-987EEA909303
device: /dev/sda
unit: sectors
first-lba: 34
last-lba: 1073741790

/dev/sda1 : start=        2048, size=     1021952, type=DE94BBA4-06D1-4D40-A16A-BFD50179D6AC, uuid=081D3F0B-1B52-4643-8A24-684A7C1A8429, name="Basic data partition", attrs="RequiredPartition GUID:63"
/dev/sda2 : start=     1024000, size=      204800, type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, uuid=607139A0-16C8-4261-BB87-253F40D707BD, name="EFI system partition", attrs="GUID:63"
/dev/sda3 : start=     1228800, size=       32768, type=E3C9E316-0B5C-4DB8-817D-F92DF00215AE, uuid=485F8027-C27D-4519-AA08-3980851152A7, name="Microsoft reserved partition", attrs="GUID:63"
/dev/sda4 : start=     1261568, size=   233179648, type=EBD0A0A2-B9E5-4433-87C0-68B6B72699C7, uuid=5BB64B2B-6646-419B-BE8A-13BB2269F0E8, name="Basic data partition"

The layout and UUIDs match the one you posted here on May 27, 2019, 7:29 PM. It’s very important exactly which image and partition layout we are talking about. I thought we were still discussing the other one with three partitions you mentioned on June 27, 2019, 3:32 PM. I just say this because if we confuse things here we won’t be able to properly help you. So in any case you want support with this, make sure you give us all the details for each image you want us to help you with. Every partition layout is different and there is no general answer or solution for all of them.

Ok, back to the partition layout from 27th of May. The d1.mbr file you uploaded has four partitions and the last partitions starts at sector 1261568 and is 233179648 sectors in size (see numbers above). Now that makes 234441216 the least amount of sectors you need on a disk to be able to deploy this image. 233179648 * 512 KB sector size is roughly 112 GB and it won’t fit onto a 60 GB disk! Too bad I can’t fix this d1.mbr for you because I am missing some important information.

In that post mentioned we see that the partitions 2, 3 and 4 are detected as fixed size and therefore FOG does not shrink those when capturing the image. So at this point we don’t know how much data you have on that partition and can’t fix the d1.mbr.

To be able to fix this we would need to find out why FOG detects the fourth partitions as fixed. Can you please schedule a debug upload task on the machine you usually capture the image from. Boot it up and when you get to the shell run parted /dev/sda print, take a picture and post here.

[Matthieu Jacquart]

@Sebastian-Roth Hi, to speak about the same and only d1.mbr, d1.partitions and d1.miminumpartitions, I continue test and I made fresh win10 UEFI install and revert Fog to 1.5.5 version. But problem was the same, capture and deploy image on SSD 128GB is ok, but deploy on 50GB disk failed.
A colleague made some change on these 3 files (d1.mbr and the 2 d1 partitions files) to change dev/sda4 size from 248745984 to 104857600, and now it works great !!!
So I think it’s a Fog capture problem, here’s a link with 2 folders : Win10 for original capture which failed on smaller disk, and 50GB with working modified files
https://send.firefox.com/download/28dabd13930e0bc0/#BMGg3OazBC_kUld0vkPs6w

Thanks

[Quazz]

@Matthieu-Jacquart What’s on the first partition? It’s currently not marked as fixed size, is this to be expected?

[Matthieu Jacquart]

@Quazz Hi, first partition is Recover (500Mb), I don’t know why it’s not marked as fixed, I still can edit file ta add 1: at the beginning ? But sda1 size is not the on d1.partitions and d1.miminum.partitions

[Sebastian Roth]

@Matthieu-Jacquart Thanks for posting more details. Can you do me one more favor. On the machine you capture this image from schedule a debug capture task. Boot it up and when you get to the shell run parted -l. Take a picture of the output and post here. Make sure we have all the information in the picture - especially the flags at the end of each line.

We definitely should find out why it doesn’t mark the first partition as fixed and this will help us!

[Matthieu Jacquart]

@Sebastian-Roth Hi, sorry for the delay but so much work and so little time…
here is the result from parted -l command with debug capture task, il it can help…

I’m on vacation in few minutes and I’ll be back in August 16th

Thanks

[Sebastian Roth]

@Matthieu-Jacquart Thanks!! I’ll play with this over the weekend and will let you know what I find.

[Sebastian Roth]

@Matthieu-Jacquart I have tested a fair bit and it seems like some or most of the issues you reported here seem to be related to the problem we are tracking down in another topic as well. I am testing fixes at the moment.

But there is still one thing that doesn’t make sense for me. Your d1.fixed_size_partitions being set to :2:3 just doesn’t add up for me. My guess is that with reverting back to 1.5.5 you are still using the “old” inits that use label based detection of fixed partitions.

I will let you know as soon as the new init’s are ready so you can test!

[Sebastian Roth]

@Matthieu-Jacquart Ok, new inits are ready. Download 64 Bit and 32 Bit and put those in /var/www/html/fog/service/ipxe/ dir on your FOG server. Probably good to rename the original ones instead of overwriting.

As well you need to check the RAM SIZE option in FOG configuration, FOg settings. Need to be 275000 with the newer inits.

With those I hope you won’t have the random issues anymore.

[Matthieu Jacquart]

@Sebastian-Roth Hi, thanks so much for your work and help, I’ll test as soon as I came back to work, after 19th August.