Parent Child on ESX Hosts

[toleolu]

I was talking to the people who manage our virtual environment about defining parent child relationships on the VM's we have running. The physical boxes run ESX, and my thinking was that the physical ESX box is the parent to any VM's running on that physical box. But what they're telling me is, since those VM's actually move from one physical box to another based on load balancing, maintenance, etc. that won't work because a physical ESX box that may be the parent to a number of VM's today, may not be the parent to those same VM's tomorrow.

I was curious if that's the way it works in other shops, and if so, is there a way to manage that, or do we just have to expect that if one of the physical ESX boxes goes down, we're going to start getting alerts from all the VM's on that box.

Mahalo

[vAJ]

Charles,

This gets back to a level of active monitoring that seems to be difficult to build into XI.

I have the same concerns around a very large virtual environment we have and again with our SQL clustering where we have databases move nodes at DBA discretion.

Not a whole lot I've seen that we can do in this regard. Other, more focused products do this, but at a very high cost. One that we've used has a name similar to "celestial breezes".

-Andrew

[toleolu]

Thanks, I can see where this would be tough to keep track of, vm's jumping around from one box to another.

I'm wondering if the solution is to remove the VM's from Nagios, since according to what I'm told, they already monitor the VM's via tools built into VMware, and just monitor the ESX hosts. We're not a real big shop, I think we have about 10 or 12 ESX hosts now, so maybe in our case, just monitoring the hosts would work.

[vAJ]

I monitor both in XI environment and don't really feel the need to have that correlation in this tool.

We have a large hosting platform with multiple layers of interdependent systems. We're working on our own "customer impact" tool that takes alerts out of Nagios and checks them against our own BI warehouse where live data about which DB/web/ESX/SAN resources the customer is connected to. When things head south, we want to be able to determine quickly what the customer impact is and better gauge level of priority notifications that our operations center sends out.

If it's isolated to one webserver that maybe a hundred or so customers live on, then the response is to engage an on-call engineer and prepare to migrate customers to other resources. If it's a large % of our customer base, we wake up VPs...

[toleolu]

I went back and talked the VM guys about removing the clients on the VM's. Turns out the parent child thing isn't really that big of an issue in our case because of the way the ESX boxes are clustered, if a host goes down, the VM's are automatically moved to another available ESX host so while some alerts may go out during the transition, it should recover pretty quickly, so I guess we'll just let it go.

Guess I should focus on the switches since the guys who manage our network can't really tell me which switch a given server is plugged into. Their response was that there's a tag on the network cable plugged into the server that identifies which switch it's plugged into, but I'm not gonna go through all that spaghetti in the back of those racks looking for that.

Government IT, gotta love it.

[lmiltchev]

You can try monitoring the ESX hosts only by following the steps, outlined here:

http://assets.nagios.com/downloads/nagi ... ios_XI.pdf

Guess I should focus on the switches

Let us know if you run into some issues.

[Box293]

the guys who manage our network can't really tell me which switch a given server is plugged into

If your switches have Cisco Discovery Protocal (CDP) enabled then you will be able to identify the switch/port from the vSphere Client.

I don't have a screenshot I can give you but I can explain how you can see it:

• Open vSphere Client
  Inventory > Hosts and Clusters
  Select the ESXi host
  Click the Configuration tab
  Under Hardware click Networking
  You will have a list of vSwitches
  To the right of a vSwitch is a list of Physical Adapters
  Each Adapter has a speach bubble next to it
  Click the speach bubble and it will tell you the info about the switch it is connected to, it will be something like Ge1/0/23

[toleolu]

Thanks, unfortunately I don't have access to that in VSphere, I'm pretty much locked down to controlling the VM's I have access to.

When I said focus on the switches, I meant the switches the rest of our physical servers are plugged into. There's a bunch of smart Cisco switches in there, but for some reason, the network guys can't seem to tell me what's plugged into them. Which I don't understand that, but then I'm not a Cisco guy, so I have to go by what the network team tells me.

No worries, like I said, "Government IT". What are ya gonna do????

[vAJ]

They obviously haven't kept up with good switch management, which includes labeling your ports.

If they were labeled in the switch config, you'd be able to see all of that with a simple SNMP query.

[tmcdonald]

@toleolu, is it safe to lock this thread up? It's been a week so just checking in.