Differences

This shows you the differences between two versions of the page.

--- services:fog [2017/08/23 19:50]
djgalloway
+++ services:fog [2020/11/13 22:17]
djgalloway [Adding a new distro]
@@ Line 1: / Line 1: @@
-====== FOG - WIP ======
+====== FOG ======
 ===== Summary =====
-We're currently testing [[https://fogproject.org/|FOG]].  These are just some quick how-tos and notes.
+[[https://fogproject.org/|FOG]] is in use in the Sepia lab.  It enables us to reimage baremetal testnodes before every job.
-The deployment of a FOG server could be mostly automated via ansible.  I wrote a playbook to take the ansible inventory and make it into a FOG-consumable CSV but it still has to be manually uploaded via the web UI.
+Our instance is currently hosted at http://fog.front.sepia.ceph.com/fog.
-Zack is currently working on a teuthology branch that'll allow us to run a proof-of-concept teuthology run where the testnodes got reimaged after every job.
+**Username:** fog\\
+**Password:** Standard root password
-Our instance is currently hosted at http://172.21.13.245/fog.  Login is fog:password.
 ===== How-Tos =====
-==== Capturing an OS image ====
+==== Adding a new distro ====
-Each testnode machine type will need an image for each distro version.  e.g., mira_ubuntu_16.04, smithi_ubuntu_14.04, etc.
-To capture an image:
+The [[https://github.com/ceph/ceph-build/pull/1706|Jenkins job]] does this automatically now.
-  - Create the image in FOG
+==== Capturing OS images ====
-    - http://172.21.13.245/fog/management/index.php?node=image&sub=add
+This can be done manually by basically deciphering the bash monster in https://github.com/ceph/ceph-build/tree/master/sepia-fog-images
-    - Name should be in a $MACHINETYPE_$OS_$VERSION format
-      - e.g., mira_ubuntu_16.04, smithi_ubuntu_14.04, etc.
-    - Set **Operating System** to Linux
-    - The rest of the defaults are fine
-  - Reimage a testnode with the OS image you want to capture
-    - Do this using [[:testnodereimage|cobbler]]
-  - Once the testnode has been reimaged and you've confirmed ceph-cm-ansible ran successfully, update DHCP
-    - ''ssh store01.front.sepia.ceph.com''
-    - ''sudo -i''
-    - Edit ''/etc/dhcp/dhcpd.front.conf''
-    - Find the entry for the testnode you just reimaged and add<code>
-next-server 172.21.13.245;
-filename "/undionly.kpxe";
-# Example:
+  - Navigate to https://jenkins.ceph.com/job/sepia-fog-images/build?delay=0sec
-    host mira082 {
+  - Choose which machine types and distros you want to capture images for
-      hardware ethernet 00:25:90:09:e2:0a;
+  - Click **Build**
-      fixed-address 172.21.7.120;
+  - ...
-      next-server 172.21.13.245;
+  - Profit
-      filename "/undionly.kpxe";
-    }
-</code>
-  - Restart dhcpd
-    - ''service dhcpd restart''
-  - Back in FOG, open the host entry (Host Management -> search for hostname)
-  - Set the **Host Image** to the image you want to capture
-  - Click **UPDATE**
-  - On the left sidebar, click **Basic Tasks**
-  - Click **Capture**
-  - Now power cycle the testnode
-The testnode will reboot and PXE boot to the FOG server.  It will automatically run its LiveOS that will capture the image and upload it to FOG.
+If capturing any image fails, the job is configured to cancel the OS capture and will leave the testnodes locked so you can debug/investigate.
-Repeat this process for each distro and machine type you want to capture an image of.  Be sure to comment the changes in the DHCP config each time so you can PXE boot and reimage via Cobbler.

Sepia Lab Wiki

User Tools

Site Tools

Differences

Page Tools