Network Management

Hi All,

I'm new to AW, as well as Linux.  So I'm totally baffled by these out of memory error messages I see on AW console: 

I installed AW about a month ago, using an .iso for installing it on ESXi 4.1 running on a Dell PE T610 server.  AW lists version 8.2.5.  My ESXi config has 6GB memory, 40GB disk.  I only have about 8 devices added to AW so far, so why would there be any memory issues at this early stage?

TIA,

Mike

And below is the Resource Allocation tab data for this virtual AW server on ESXi 4.1:

Anybody?

Hi Mike,

We need to reserver the resources to Airwave instance, I see from screen shot memory is not reserverd to this instance.

We do not recommand shared memory, try reserver the resources and check the status.

Regards,

Pavan

Hi Pavan,

I'm not sure what you mean by "reserver the resources to Airwave instance"?

AW AMP was installed in an ESXi 4.1 environment, with 6GB memory, 40GB HDD allocated to it.

On my screen shot I do see mention of "Reservation" under MEMORY and CPU sections.  Are you saying I have to configure my AW AMP ESXi host to "reserve" a certain amount of memory?  I have allocated 6GB of memory, but I guess "reserving" means something else?  I would have thought memory usage was like disk usage - a certain amount of space is allocated, and the guest (AW AMP) uses disk and memory as needed, up to the allocated amount?

Thanks,

Mike

Hi

Reserve nothing but memory which we allocated must to dedicated to airwave instance itself, it should not be used by any other machine which is installed on VM host.

Regards,

Pavan

At minimum, the smallest install specs should be:

4 core, 8GB RAM, 80GB disk.

It needs the RAM to start up the daemons in the right timing, otherwise the initialization scripts that deploy the db and feature templates fails to start.  If you go under the support menu - launch the diagnostics tarball, and then download it - the service_watcher and low_level_service_watcher will show which items are attempting to gather memory needed.

Ive never configured a VMware virtual machine by "reserving" resources.  All I ever do is allocate a certain amount of memory and disk space and let VMware take care of the rest.  Do you have any article or FAQ you can share with me that describes exactly how Airwave needs to be configured when running as an ESXi virtual machine?

Loading the AirWave install ova observes the specs I provided.  The ISO will only give you the amount of space that was allocated when the VM was initially setup / allocated.

If you're manually deploying a VM to use the ISO, assuming VM 6.5 -> here are the steps:

1. Highlight the 'host'

2. Create/Register VM

3. Create a new VM

4. Input name, OS=linux, OS family=Cent 4/5+

5. Choose datastore

6. @ Customize settings: 4 CPU, 8 GB RAM, 80 GB disk (or more), then mount ISO

I'll try increasing RAM from 6GB to 8GB and see if the memory errors go away.  I'm monitoring a very small number of devices, probably less than 25 when I'm done.  So I imagine the 80GB is not needed - that's why I picked 40GB.

What is the "reservation" that Pavan suggested?  Is there some special reservation process for memory for a given virtual machine in ESXi that I'm supposed to do?  Or was he just suggesting to allocate memory in the properties of the vm?

I think Pavan is referring to if you're using a dedicated VM for AirWave or if the VM has other instances running.  The issue will be disk I/O contention.  AirWave being a monitoring system is heavy on read/writes.  Having the incorrect allocation of memory affects the performance directly.

Hi Rob,

I'm not sure I follow:  I have an ESXi 4.1 host running 5 VMs.  4 of the VMs are various MS Windows servers.  The 5th VM is dedicated to AirWave (8.2.5 I think), which I installed using an ISO I had downloaded from Aruba.  How could this one VM be running multiple instances?  I guess I don't know VMware as well as I thought :)

Initially my AW usage will be very minimal, with probably no more than 10 to 15 devices as I test AW in my office.  I have a single Aruba AP 303H, which has maybe 5 to 10 wifi clients.  Then there's a few Cisco and Netgear switches, but not much else.  Are there any guidelines as to how much memory an AW VM needs for extremely small environments?  I assume the more devices one monitors with AW, the more traffic and data to be stored.  So I would assume 80GB HDD is really not needed for small installs?  And the 8GB memory?

VMware does support "reserving" memory - which I have never needed to do.  Is Pavin suggesting editing my AW VM's properties and making a memory "reservation"?  I imagine that reserving memory might reserve a chunk of physical non-shared memory for the VM - resulting in faster access since no paging to disk needs to be done for the reserved memory?  This is all just a big guess, as I have very little experience in tuning VMware...

Thanks,

Mike

When I say multiple instances, I'm referring to the VM Host.  So your host has 5 instances.

You can probably last off of 40-50 GB of disk, the OS itself takes about 10GB.  My test setups that live for about 2 weeks at a time take up about 25GB each.  So if you use less disk, then you might consider lowering the amount of data maintained on the AMP from AMP Setup -> General -> Historical Data Retention.

8GB RAM should be the minimum.  Otherwise you'll see services flapping during the initial start up.

As far as VM reservation, the 8GB allocated to the instance should be enough.

Oh, so some people go to the trouble of installing VMware and only install one VM - for AirWave?  Seems like overkill, and sort of a waste of a machine, *if* you have a powerful server...

I don't get the out of error messages during startup.  It happens many hours later (maybe even a day or two).  Do the error messages I posted earlier help you identify the problem?

Thanks,

Mike

No, it's more like a customer puts AirWave on a VM Host that is already heavily loaded with other equally or more important infrastructure.  It's still just 1 instance of AirWave, but if it's on a server that also hosts another database instance, then the 2 instances require a lot of IOPs.  When it reaches disk contention, 1 instance gets a higher priority to use the disk than the other, in which case you'll see performance trade off.