Node info is correct but no VMs show up.

[lwimble]

I could use a little guidance on where to go from the point I have reached in installing and configuring Opennode 6.0 beta 1. 

First,  a little background: While I've been running Linux systems for many years,  I am quite new to OpenVZ and this type of virtualization.  I've never used certmaster or func, and I am definitely NOT a python programmer (I write in C; Python looks like gibberish to me).

I installed CentOS 6.2 from a CentOS mirror, and ran the conversion script to opennode 6.0 beta 1 (By the way, this was a largely painless process except for the fact that I had to download and install libploop manually and re-run portions of the conversion script).  In any case, after as LOT of googling and learning about certmaster, I finally have my node showing up in the ONC. 

My setup is something like this: The node machine is named "drosera", IP 192.168.101.179.  It currently has one virtual machine running on it called "oms", IP 192.168.101.177.  One thing about this machine: eth0 is inoperative.  Everything is running on eth1.

In any case, I can see "drosera 0.0.0.0/32" in the panel on the left side of the ONC.    When I click on it, I get a tab with all kinds of info including memory used, disk used, etc.  All of this information is correct, so I know the two systems are talking to one another. 

The problem is that no information shows up under the sub-tabs entitled "VMs", "Network", or "Templates".  In the "Shell tab", I see a red flashing cursor with a square that says "Connect" which sits there forever in this state.  Why doesn't the "OMS" vm show up in the list?

Finally, I have a question about the OMS shell: What is this useful for?  Clearly it's a hierarchical filesystem of some kind, but what does it do?  Also, how to I get rid of hosts in the list which aren't a part of my network like testvirt5.active.ee ?

Thanks In Advance,
Larry Wimble

[activesys]

centos6-to-opennode6 script might be a bit outdated sometimes - but ploop packages are now in repo - so that should be ok now.

Could you enter into OMS and update OMS with /opt/oms/update.sh ?
And then see if problems still exist.

[lwimble]

I found the problem.  It had to do with an incompatibility between libvirt and OpenVZ.  It's described here: http://comments.gmane.org/gmane.comp.emulators.libvirt.user/2620

This makes sense that I would see the system metrics but no information on anything having to do with OpenVZ.

I downloaded the latest version of libvirt, patched it, then compiled and installed it.  Restarted everything.  OMS works great!

The only question I am left with is this: How do I delete hosts from the left side list in the ONC?  There should only be one host there (drosera), yet testvirt5.active.ee shows up and I know of no way to delete it.  The only place any reference to this host shows up on the entire system is in /opt/oms/db/data.fs.  There are no certs or csrs for this host anywhere.

Thoughts?

Thanks,
Larry

[ilja_l]

Hi, Larry

some comments:

1) We've patched libvirt to correctly handle new vzctl, so please, do yum update libvirtd. A bug + patch are also submitted upstream, but I guess it'll take some time.

2) Regarding extra host: well, one option is to use OMS shell (this also covers your question from the first entry):

Code: [Select]

cd computes/
ls -l 
rm <UID>

Also, check that cermaster-ca --list-signed doesn't contain testvirt5 entry. If it does, just do certmaster-ca -c testvirt5.active.ee. I suspect you've downloaded OMS exactly in the time span when it was a bit broken -- cleanup script wasn't working correctly on the OMS VM.

3) Coming back a bit to OMS shell: our initial thought was that we need a 'Cisco IOS' kind of shell for managing the infrastructure - so this is essentially what OMS Shell is about (btw, you can connect to it directly with SSH: by default it runs on port 6022 in OMS VM). Paths in OMS shell also correspond directly to REST paths, which is actually being used by the Opennode Console (aka ONC). You can get a bit more information about the idea behind here: http://opennodecloud.com/docs/opennode.oms.core/developers.html .

cheers,
Ilya

[ilja_l]

Ah, another comment

The only question I am left with is this: How do I delete hosts from the left side list in the ONC?  There should only be one host there (drosera), yet testvirt5.active.ee shows up and I know of no way to delete it.  The only place any reference to this host shows up on the entire system is in /opt/oms/db/data.fs.  There are no certs or csrs for this host anywhere.

To explain why testvirt5 is there, just a bit of how explanation how OMS works: the design of OMS supports deletion of DB at any point. DB is repopulated by a 'sync' daemon (you can see all the daemons in oms shell: ps ax. 'kill -STOP sync' to pause it). Now, syncing works by examining all of the registered hosts (certmaster db) and issuing calls to a our agent (currently func based - most probably will change in future) about the capabilities etc of the underlying system. So the only 'persistent' part is CSR/CRT files in /etc/pki/certmaster - and I guess the certs were simply left there - as you have probably guessed, they are from the testbed machine.

Ilya

[lwimble]

Ah, using the OMS shell worked fine.  It never occurred to me that you could get a "recursive" delete using just "rm".     

Everything seems to be working at this point, albeit I am left with one final question.  When I went to create an OpenVZ virtual machine, it created one but it didn't use the parameters I had set in terms of memory, disk, etc.  it also didn't start the VM.  I had to start it manually with vzctl.  Any thoughts here?

Also, is it possible to edit those settings once a VM is created?  I didn't see a way.

[lwimble]

...and of course, one more.....

If you're planning to migrate away from func, what are you planning to migrate to?  Just curious...

Larry

[ilja_l]

Hi, Larry

Ah, using the OMS shell worked fine.  It never occurred to me that you could get a "recursive" delete using just "rm".

     

Not sure what you mean by a 'recursive delete' - rm removes a certain Server (aka Compute). It is somewhat natural that once the top object is gone, also its children are removed. In a 'file system' world most of the objects are equal and rm should be explicit.

Everything seems to be working at this point, albeit I am left with one final question.  When I went to create an OpenVZ virtual machine, it created one but it didn't use the parameters I had set in terms of memory, disk, etc.  it also didn't start the VM.  I had to start it manually with vzctl.  Any thoughts here?

Also, is it possible to edit those settings once a VM is created?  I didn't see a way.

Do you mean that you've created it via ONC? Or in TUI? Actually, in both cases settings should propagate, if not - we'll fix that asap. Editing is currently supported in TUI (opennode -> manage -> VMs -> <pick a VM> -> edit), ONC support is in development.

If you're planning to migrate away from func, what are you planning to migrate to?  Just curious...

That's a good question - there are several candidates (e.g. http://salt.readthedocs.org/en/latest/), including writing our own agent (in Go). Basically, we have a number of typical actions scripted in Python and it would be nice not to waste that. However if we go for a change in agent, it could as well be something that works on multiple platforms without installation of additional runtimes. If you have suggestions, we'd be very happy to hear them!

PS. There were several pretty significant updates to OMS stability today, I suggest running /opt/oms/update.sh in your OMS VM.

Ilya

[lwimble]

Not sure what you mean by a 'recursive delete' - rm removes a certain Server (aka Compute). It is somewhat natural that once the top object is gone, also its children are removed. In a 'file system' world most of the objects are equal and rm should be explicit.

What I meant was that in the world of filesystems, one would use "rm -r" to recursively delete.  In the world of the OMS shell, simply deleting a node deletes everything under it without an explicit parameter to do so.  Since the "help" for "rm" didn't include any reference to a "-r" parameter, I didn't think it would do that, and frankly didn't even try.

Do you mean that you've created it via ONC? Or in TUI? Actually, in both cases settings should propagate, if not - we'll fix that asap. Editing is currently supported in TUI (opennode -> manage -> VMs -> <pick a VM> -> edit), ONC support is in development.

I created it in the ONC.  I'll experiment with this a bit more and see if I can reproduce it.

One request: Both the hardware nodes and the CTs show up in the list on the left in the ONC.  It would be nice if they either showed up in a hierarchy (the CTs indented under the hardware node) or at a minimum, the CTs had a different icon to differentiate them from a hardware node.

Finally, I want to thank you for your excellent and responsive support.... You guys make a really great product!


Larry

[ilja_l]

Hi, Larry

first of all, thanks for the kind words Having a feedback, especially positive, is very motivating. We are in active development mode, so stay tuned for updates.

About rm confusion: although we do try to make OMS shell to resemble bash/zsh as much as possible (btw, check out the embedded 'emacs'-like editor: edit <object> ), in some cases it's a bit dangerous, because POSIX semantics is pretty different. On the other hand, as it's causing a confusion, we should probably add explicit confirmation for 'recursive' deletions.

About sorting: if you have the latest OMC/ONC (/opt/oms/update.sh), then you should be able to sort machines out by appending a filter virt:yes and virt:no. We have a better indexing system in place (see /search/by-tags in oms shell), but it hasn't yet propagated to ONC.

cheers,
Ilya