CentOS provision using default template results in anaconda error populating transaction

Problem:
I’m new to The Foreman. I’m trying to Provision a CentOS machine using PXE boot. I have the client see the TFTP server, download the kickstart file (which is default but have tried edits) start the install. It is using a grub2 uefi installer and I get to the stage where anaconda reports “Starting package installation process” and it errors with 10 “Error populating transaction, anaconda is retying”. I’ve tried 3 different mirrors as the installation media. Here is the installation media entry. http://mirror.internode.on.net/pub/centos/$version/os/$arch

In a shell on the machine here is the first line of /var/run/install/cmdline.c/80-kickstart.conf on the client
inst.repo=http://mirror.internode.on.net/pub/centos/7/os/x86_64

Foreman and Proxy versions:
Version 1.16.2

Any help would be greatly appreciated.

This is an error during transfer, your connection is not good and files are getting corrupted. Download the installation DVD, do MD5SUM to verify, extract it to a HTTP local server and add installation media from DVD into Foreman, install from there that should do it until you figure out your connection.

Thanks very much for the help. I will try that however, I will be surprised if that is the issue as we have a very good internet link and I’ve tried 3 different mirrors after seeing results on other machines from fastest mirror. I’ve also bee able to do an Ubuntu netboot on the same client. Is there any other possible explanation?

Cheers

Trent

This simply works ™ I am sure there are multiple Foreman instances in the wild installing several CentOS machines as we speak :slight_smile:

All I was able to google out was error in metadata.

One idea is that UEFI is not that battle-tested as BIOS, can you try switching over to PXELinux and BIOS just to rule this out? But once Anaconda is up, it should not matter if you do UEFI or BIOS I guess.

Thanks again for the help. I tried another mirror this time over https and it worked. So you were right about the media issue.

It could be an error in metadata which was distributed across mirrors for some time window. It rarely happens.