Community
- Command Central
- MVP Program
- Monthly Mission
- Blogs
- Groups
- Events
- Media Vault
Products
- Observability
- Network Management
- Application Management
- IT Security
- IT Service Management
- System Management
- Database Management
Content Exchange
- SolarWinds Platform
- Server & Application Monitor
- Database Performance Analyzer
- Server Configuration Monitor
- Network Performance Monitor
- Network Configuration Manager
- SQL Sentry
- Web Help Desk
Free Tools & Trials
Store

Running APM 3.0 on dedicated server?

Folks,

I inherited an existing NPM 9.5 SP4 installation and I've spent my 4 days here looking at a slew of issues, most of which seem to involve APM 3.0.

The system is:

Network Elements 7575
Nodes 1607
Interfaces 3966
Volumes 2002
Alerts 213
Events 663909
Pollers 12697
Polling Engines 2

Licensed Component Monitors: 4172

They have NPM, NTA, VOiP, and APM running on the primary NPM server. There are 5855 elements on THAT server, with the remaining 1720 on the 2nd Poller.

At this point, APM is so hammered, selecting the tab yields a timeout to the website. Drilling to a node, the Applications Panel shows this error:

The "Applications" resource threw an exception..
Exception has been thrown by the target of an invocation.

I have a case open with Support, and applied a HotFix this morning...no joy. A new Diagnostics dump following this has been sent. In looking at the Requirements section of the APM Admin Guide, I see that 5-8GB of RAM is recommended for > 1000 Component Monitors.

The main box is a 3.0 GHz Xeon with 4GB....running all those modules and elements listed above. So, I think it would help to either throw a lot more RAM on that box OR move APM to another server with NOTHING else on it.

QUESTION: Is that possible? (To tilt up a dedicated APM 3.0 server.)

SECOND QUESTION: How do I Unmanage or otherwise greatly reduce the Component Monitors temporarily to test for effect on the existing server. My thought is that if I only have a handful of CM's running, I can verify the RAM issue by creeping back up until I hit a break point.

Any feedback, advice, etc. is appreciated!

Find more posts tagged with

Accepted answers

All comments

RogerWong

Hi NinjaNerd56,

1. You can't install Orion APM without Orion NPM, but more RAM will help.

Also, what operating system is Orion NPM/APM hosted on? If it's Windows 2008, consider a Windows 2003 server instead (there's a Microsoft bug in Windows 2008 regarding WMI that can make the entire system very sluggish).

2. Yes, you can unmanage components and applications to temporarily lower your component count:

Unmanaging Applications

Unmanaging Applications

If you need to deactivate an assigned application monitor for a period of time, you can do so by unmanaging them. Changes you make to the managed or unmanaged status of an application take affect after the next polling cycle.

To unmanage an application:

1. Click APM Settings > Manage Assigned Application Monitors.

2. Select the monitor and then click Unmanage.

3. Set the duration of the unmanaged state, and then click OK.

-or-

1. Click an assigned application monitor in any resource to view its APM Application Details page.

2. Click Unmanage in the Application Details section.

3. Set the duration of the unmanaged state, and then click OK.

Note: Unmanaging a node in Orion NPM automatically unmanages its assigned Orion APM applications for the same duration.

To remanage an application:

1. Click APM Settings > Manage Assigned Application Monitors.

2. Select the monitor and then click Remanage.

-or-

1. Click an assigned application monitor in any resource to view its APM Application Details page.

2. Click Remanage in the Application Details section.

NinjaNerd56

Roger,

Thanks for verifying what I suspected from reading the Admin Guide for Question #1.

We are running Server 2003 Standard, so we're limited to 4GB there. We can move to x64 since that's probably cheaper than buying another Poller.

The system is SO SLOW that I timeout frequently, so it may take me several iterations to unmanage enough instances to see an effect.

Thanks again...will post an update when finished!

FormerMember

this is a very intresting post... i have similar problems... i have orion, and then 4 additional pollers. 2 of the additional pollers run the APM module to help take the pressure off of the main poller.

I assume Orion would not benefit from a 64-bit operation system for its run-time, but never thought about the Solarwinds binaries needing more than the 4GB of RAM.

I have SQL running on a windows 2003 advanced cluster with 16GB of RAM, but the pollers all run on windows 2003 standard, with 4GB of RAM....

what do you think solarwinds?

NinjaNerd56

The update is that the NetPerfMon DB is a nice golden brown....as in toast.

I was unable to unmanage a single Application Component, much less the 4172 configured.

We're having a meeting to discuss rebuilding the entire system with sufficient resources on Monday.

I've asked our excellent SW Support contact on the open case for input on machines and resources, but would LOVE to hear from any and everybody out here with your thoughts and recommendations.

Thanks in advance!

NinjaNerd56

UPDATE:

Rebuilt the primary NPM Server with Server2008 x64, 8GB of RAM.

The SQL Server was rebuilt, also Server2008, x64, and 8GB of RAM. SQL Server 2005 with NO other DBs on this machine. (Previous installation was sharing the server with 6 other DBs).

Kept the existing 2nd Poller and Additional Web Server machines (Server2003, x32, 4GB), but did clean installs of Orion, pointing to the new SQL Server instance, etc.

The entire system is MUCH faster overall. Initial user (and management) response is 100% positive.

I have seen that the 2nd Poller is a little "stressed" after rediscovering and populating the system. I have recommended we go ahead and rebuild it with Server2008, x64, 8GB like the others. The Web Server is fine, so it will be the only Server2003, x32 system left when we're done.

Did some WireShark captures, before and after, to examine the system throughput from a network and application response time perspective. The "test" was a) Login to the Default page, wait until the page and custom map were completely drawn and b) navigate to NetFlow Traffic Analysis and wait until that page and the Top N pie charts were complete.

The "old" system took 80-90 seconds to perform those 2 steps. The rebuilt system takes 40-45. Decent results. May improve a little more after the 2nd Poller rebuild.

FWIW......