1 files changed, 36 insertions, 0 deletions
diff --git a/blog/gentoo-on-gcloud.org b/blog/gentoo-on-gcloud.org
new file mode 100644
index 0000000..d826e53
--- /dev/null
+++ b/blog/gentoo-on-gcloud.org
@@ -0,0 +1,36 @@
+* gentoo on google cloud
+
+Google Cloud only provides Debian and a few other operating system images by default, but does allow bringing your own ISO.
+
+** gentoo official qcow2 images
+
+The first thing I tried was booting one of Gentoo's official qcow2 images directly.
+
+Importing custom images requires downloading the image to a "bucket" which is basically Google's s3 object store.
+
+Downloading images to a bucket requires transfering data via a URL list. The URL list is a file hosted somewhere with a list of URLs in it, one per line. Maybe using a pastebin for the URL list could work?
+
+After getting the image downloaded to the bucket, you must go to the "Migrate to Virtual Machines" entry, which is underneath "Compute Engine". Then there will be a button that says "Import Images" (note there is another button called "Import Machine Images" which won't work). The rest should be easy.
+
+After getting the image imported, you can use it like any other gcloud disk image.
+
+After doing these steps, I created a new vm instance with the Gentoo qcow2 image as the base, but it failed to boot with an error saying something along the lines of "no bootable media found".
+
+I am not exactly sure what is wrong with gcloud, but these images boot fine in qemu for me.
+
+** hijacking a debian image
+Instead of using one of the official Gentoo images, we decided to "hijack" one of gclouds already booting Debian images.
+
+To do this, I created a new instance using the default Debian boot disk, and then cloned the instance's disk while the instance was shut down. Then I booted the instance back up, and attached the clone'd disk to the instance.
+
+From here, I made a brand new filesystem on the main rootfs partition, but left alone the rest of the disk. You can install Gentoo from a stage3 and chroot in like normal to configure stuff, or install however you'd like.
+
+When installing the kernel, I just deleted the old debian ~/boot~ contents, and installed ~gentoo-kernel-bin~, and ran ~grub-mkconfig~ like you'd normally do. This also failed to boot, with an error about $UUID not existing. It looks like the old EFI binaries/bootloaders are somehow conflicting with out setup.
+
+The trick that got it to boot, was clearing out the old EFI binaries completely, and then installing grub with the ~--removable~ flag. Without the ~--removable~ flag, it would not boot.
+
+After installing our new bootloader, I shut down the instance, detached the boot disk, and reattached the Gentoo disk as the new boot disk. Then booted into Gentoo!
+
+* tldr
+
+To get it to boot, you hijack a working image by making a new rootfs on the rootfs partition, replace the kernel, initramfs and then the bootloader with a "removable" bootloader (aka ~grub-install --removable~).