Kernel update - Storage Nodes

[Wayne Workman]

The kernel update process (within FOG Configuration) does not also place a copy of the newly downloaded kernel onto all storage nodes, which means computers network booting from those nodes do not have the newly downloaded kernel available.

I’m not sure if there is an easy solution to this.

For the moment, we are just downloading the kernels normally via Kernel Update, and then using scp to move them to the nodes.

[JJ Fullmer]

This sounds more like a feature that doesn’t exist than a bug. Just saying
I mean what if you update the kernel in one node but then that kernel has a problem. Wouldn’t you want the stable kernel to stay put on the other nodes?
Or maybe you have a custom kernel on one node but not any others?

Granted it sounds like a useful feature, updating the master node causes all nodes to update, that would be cool. Doesn’t sound all that easy though.

[Wayne Workman]

Moved to feature requests.

[Sebastian Roth]

@Wayne-Workman I am with Arrowhead here. Don’t think you want auto-update the kernel on all nodes. Why would you?

[Quazz]

@Sebastian-Roth Perhaps an option to enable it would be a good idea? I can see it being useful in environments that are very uniform or have no compatibility issues.

[Tom Elliott]

@Quazz Maybe I’ve spoiled myself, but why not just rerun the installer? It will automatically load the current kernels to the system.

Building a new system to sync kernels is doable, but not very simple.

[Quazz]

@Tom-Elliott Yeah, I get it, but people will forever be lazy

[Wayne Workman]

I’m not even asking for syncing. And maybe I don’t need the latest kernel? Maybe I need an old one? All I’m asking is that the kernel update process cycle through all the storage nodes and put what you downloaded on them all.

[JJ Fullmer]

@Wayne-Workman Couldn’t this just be locally scripted? I mean personally if I update fog on my main server I run the update on my node too. But I only have one node, so it’s no big deal. But if you have a script to update fog you could probably have one script that runs it on each node after the first one finishes. I don’t know off the top of my head the command for it, but I’m pretty sure you can do it with an ssh command in a script. Would need to deal with the ssh key setup and stuff first, but it can be done.
I suppose that since fog should already have some access to the storage nodes, the fog installer could say hey I’m updating, and look at the database and say hey look storage nodes and force a git pull or svn up on each and run the installer with -y on each one.

So it would take some effort but I’m pretty sure it can be made into a script. I think a self made script would be better, because eventually Fog will be a stable release without so many updates.

Of course if you were only updating the kernel and not the whole installation, adding an optional scp or rsync command to all storage nodes after updating on the main server would be fairly easy I imagine.

[Sebastian Roth]

Well, I’d use clusterssh for that I guess. Scripting would be good as well but with clusterssh you see what is happening on all the storage nodes. Don’t bother about kernel updating via the web gui but do it all at once on all your nodes with clusterssh!

[Wayne Workman]

@Arrowhead-IT Of course I can script it. But not everyone is familiar enough with whats involved to do that. My intention is to eliminate extra manual work in the back-end involving FOG. I’ll see about writing a patch.