Multicasting Swap Partition Problem

[cspence]

I’ve been having trouble with multicasting with a Multiple Partition - Single Disk image with a swap partition. So here are some specifications I have before getting into this.

[B][U]Specifications[/U][/B]
[U]FOG Server[/U]
OS: Debian 7
IP: 10.2.0.50/24
[COLOR=#ff0000]FOG Version: 1.2.0 (Edit: Thanks for pointing out I left it off, Wayne Workman.)[/COLOR]
Locally Hosted Services: TFTP, NFS
Firewall: Disabled (not in production)

[U]DHCP Server[/U]
OS: Windows Server 2008 R2
DHCP Options configured to use 10.2.0.50 as the TFTP server as directed

[U]Example Targets[/U]
IPs: 10.1.5.1/24, 10.1.5.1/24
Hard Drive Space: 250GB

[U]Test Image[/U]
OS: Debian 7
Size: 1.2GB
Partitioning Method: MBR
Partition 1: Primary, Bootable, /
Partition 2: Primary, Swap, swapon
Bootloader: GRUB

[B][U]Test Scenario 1[/U][/B]
Description: Multicasting Multiple Partition - Single Disk Image (fstab Config 1)
Special note: The fstab used UUIDs to identify all partitions.

[U]Issues[/U]
Deployment would successfully send the root partition and freeze on the swap partition. To get around it, I would have to stop the tasks and hard reboot the machines. The images were bootable.

Inspecting the machine showed that the swap partition was created, but that they had not been activated (UUID mismatch issue - Example thread: [url]http://fogproject.org/forum/threads/imaging-linux-systems-uuid-for-swap-not-matching-on-deployed-systems-eh.10901/[/url]).

[B][U]Test Scenario 2[/U][/B]
Description: Multicasting Single Disk Image
Special note: The fstab used UUIDs to identify all partitions.

[U]Issues[/U]
Imaging successful. No hangs during imaging.

[B][U]Test Scenario 3[/U][/B]
Description: Multicasting Multiple Partition - Single Disk Image (fstab Config 2)
Special note: The fstab used /dev/sda* format to identify all partitions.

[U]Issues[/U]
Again, deployment would successfully send the root partition and freeze on the swap partition. To get around it, I would have to stop the tasks and hard reboot the machines. The images were bootable.

The swap partition was created, formatted properly, and was active.

[B][U]Concluding Comments & Questions[/U][/B]
For the record, I had the same thing happen when I had my root partition as a primary and the swap as a logical partition.

The swap deployment hang issues only occur when using Multicast jobs. Unicast deployment does not hang and the job completes without intervening by powering off the machines and cancelling the tasks.

Is this a known issue or is this just something I’m doing?

[Wayne Workman]

I can’t lie, when I looked through that incredibly detailed post, trying to find the FOG version and realizing it wasn’t there, I smiled.

[cspence]

[quote=“Wayne Workman, post: 42849, member: 28155”]I can’t lie, when I looked through that incredibly detailed post, trying to find the FOG version and realizing it wasn’t there, I smiled. :)[/quote]

I rolled my eyes when I saw your reply. I edited it into the original post and gave you credit.

I forget that not everyone runs the latest version.

[Sebastian Roth]

Could something be wrong with that particular swap partition?? Maybe try recreate swap on the client before uploading an image?! Just a wild guess. Will think about the issue a little more…

[cspence]

[quote=“Uncle Frank, post: 42863, member: 28116”]Could something be wrong with that particular swap partition?? Maybe try recreate swap on the client before uploading an image?! Just a wild guess. Will think about the issue a little more…[/quote]

I also noticed that partimage does not officially support ext4. When conducting these tests initially, I used an ext4 partition for the root partition. I thought it would be best to ensure that it wasn’t a complication from using ext4 (a long shot) so I installed Debian 7 from scratch again but with an ext3 partition. I ran the same tests again and got the same results.

It’s just strange to me that unicast jobs work fine, but multicast jobs can’t handle swap partitions in multi-partition mode.

[Sebastian Roth]

Alright, sounds like this is definitely an issue with FOG multicasting. I can’t test this in our labs straight away as we have linux setup with wubi (swap is within the container file) which does not cause any problems! I see if I can setup a couple of testing machines to try and reproduce this issue…

[cspence]

Sounds good. If you need anything from my end, just get back to me. I’ll be glad to help (I have a couple of routes, several subnets, and several VLANs to use for thorough testing).

[Sebastian Roth]

I had a quick look through the shell scripts and saw that with FOG 1.2.0 there are actually differences in how swap partitions are being handled depending on how clients are uploaded/deployed (multicast vs. single download and single disk vs. multiple partitions)! As I am not a developer I can’t tell you why things are like this. Maybe Tom or Junkhacker can shed a light on this.
But from what I can see there have been major changes in those scripts in the last months. If you got a spare machine you could easily setup another testing FOG server and install the current SVN version (see here: [url]http://www.fogproject.org/wiki/index.php/SVN[/url]). I can’t promise you anything but I wouldn’t be surprised if swap is working with that… If not we can still start digging into the shell scripts to see where things go wrong.

[Junkhacker]

i think i remember something about muti-partition multicast imaging having an issue and being fixed in svn. i recommend upgrading to svn revision 2948, but have a backup you can revert to in case you have problems.
also, regarding this:
[quote=“cspence, post: 42865, member: 28749”]I also noticed that partimage does not officially support ext4. [/quote]
fog uses partclone, not partimage, for any version newer then 1.0.0

[cspence]

[quote=“Junkhacker, post: 42892, member: 21583”]i think i remember something about muti-partition multicast imaging having an issue and being fixed in svn. i recommend upgrading to svn revision 2948, but have a backup you can revert to in case you have problems.
also, regarding this:

fog uses partclone, not partimage, for any version newer then 1.0.0[/quote]

Good things to know. I’ll look into that revision and get you some feedback.

I should have confirmed what imaging system it was using now. That would explain why ext4 had no issue.

[cspence]

Now that we got the NIC issue figured out, I got a chance to test this problem out with SVN 3053.

Unfortunately at this point, it appears I need to do some more troubleshooting because the multicast session fails to start at all. Currently, I tried using Multiple Partition - Single Disk Imaging. I’ll pick this up again a bit later (hopefully this week). This week has proven to be busier than usual.

[Wolfbane8653]

Check out the multicasting wiki. It may help point out an issue.
[url]http://fogproject.org/wiki/index.php/Multicasting[/url]

Also please update to the latest revision. Currently r3077 has been issued.

[Sebastian Roth]

Yes, please upgrade. There have been a lot of bugs and fixes lately…