RHEV trouble shooting
Hi,
I'm getting to the somewhat serious stage of setting up a pilot for my company.
As I don't want to get into a situation where I can't troubleshoot, I am asking some possibly stupid questions mostly coming out of my experience with VMware.
-
If I lose the manager - can I connect to a host directly and start VMs? and do other things - say change a network etc?
-
Do VMs ever get Orhpaned? in other words, Manager can't find them, but they still exist? And if so how do you re-register them with Manager?
-
Are their any performance monitoring tools... to help me with a VM that is performing poorly - to help diagnose if it is a compute, memory, disk or network issue?
I'm not looking for someone to write a book on this, but any pointers to what I should be reading would be appreciated!
Thanks
Bill
Responses
Hi Bill
A1. Yes and No. VM operations can be carried out at the hypervisor level (using vdsClient) , but are not supported. Moreover, any operation carried outside of the RHEV Manager is not reflected in the engine database and can create an inconsistant view when later brought online. So, the short answer is really NO.
A2. I've not come across an orphaned VM in RHEV before. I would imagine it would take some backend friggery to get this back inline if it we're to come up though. On the very rare occation that I've experienced a VM in an "Unknow" state, a VM restart is required to clear.
A3. Outside of the normal performance monitoring tools available on the VM itself (iostat, vmstat, sar etc) there are Jasper reports available via the WEBUI. These are based mostly around capacity management as opposed to performance though. Basic realtime performance of your hypervisor and VM estate can be viewed via the Hosts tab on the WebUI. If you are looking for something like IOPs per VM though, I don't think there is anything. Yet (subject of RFEs)
Others may have suggestions.....
Since we don't have an easy way to "vote" on a response, this is my endorsement ;-)
regarding 1... and specifically the question about modifying network settings, that particular requirement seems challenging - mostly because using the RHEVM (and the WebUI) you are going to create a network, with a human-readable alias... yet, behind the scenes, it is simply a bond (and br). I don't actually assign an IP on the Hypervisor for my Guest VM network bonds. So - you have a bridge (using your human-readable name) with a bond (likely) and then a vnet interface for the VM.
-
I have had an experience where VMs were Orphaned (my own doing - I pushed a "standardized" /etc/sudoers from my Satellite config channel and trashed my Hypervisor). You absolutely should call Red Hat in this situation, but - it was a quick/easy fix (once they provide the data ;-) The call to Red Hat will ensure that you have backed things up, nothing else is broken behind the scenes that you may not realize...and most importantly - gives them the data they need to get you out of whatever scenario you might get yourself in to.
-
I have nothing more to add regarding reporting. I tend to rely on the basic RHEV WebUI for very basic stats and I would pursue the rhev-reports if I need to go deeper. Otherwise, standard OS commands at the CLI (I use RHEL as a Hypervisor).
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
