I spent some time yesterday going through an evaluation of the upcoming release of Microsoft Operations Manager 2007, and figured I’d share my evaluation notes with you, dear reader. I’ve used MOM before, and I wanted to see what was new.
I’m looking at SCOM 2007 because I’d run into a showstopper with UniCenter: it can’t monitor my 64-bit machines. Since we’re going into production with 64-bit SQL Server for the data warehouses, I need a short-term (6-12 month) solution that will handle the 64-bit monitoring until UniCenter has 64-bit agents ready.
Server Monitoring Features I Need
My main criteria in order of importance:
64-bit OS and Application Monitoring
I’m looking for both operating system monitoring and application statistics monitoring. It’s not enough to just see that a 64-bit system is up – I need to know application events, like when SQL Server is running out of log space for a particular database. UniCenter is able to gather this data, and so should my short-term solution.
Easy To Deploy & Maintain
This system has to reduce our workload, not increase it. My goals for deployment are:
- Monitoring system deployment in 1 day
- Install agents on each server in 1 hour
- Add & edit users and alerts in 5 minutes
- Have everyone on the Windows group able to do those functions in less than one hour of training
There are plenty of monitoring products out there that achieve the rest of my goals, but miss on this ease-of-use one. I’m intimately familiar with ServersAlive, for example, but if I’m the only guy in-house who can maintain it, and it takes me one day a week, then we’re worse off than when we started.
Simple Monitoring Dashboard
Email alerting is good, but when a whole bunch of things go wrong at once and then get fixed, like during a network outage, I want to be able to see what’s still broken at a glance. A good example of this is UniCenter’s display up in the CORE. Bonus points if that console is accessible via a PDA.
Server Performance Trending
Right now, I’m manually gathering statistics on my machines via Windows Performance Monitor one week per month. I assemble that data into Excel and run calculations on it to stay on top of my machine health. Unfortunately, that takes about a day per month to collect, report and analyze. I’d like to automate that process. (UniCenter does this.)
With these goals in mind, I evaluated Microsoft Operations Manager 2007 Release Candidate 1. The product is mostly done, but things like documentation and polish are still hit-or-miss.
SCOM 2007 Review: Positives
SCOM 2007 is 64-bit capable.
It monitors 64-bit operating systems and applications seamlessly. The agents install and work fine without a hitch. OM2007 automatically detects and installs the appropriate 32bit or 64bit agents, including SQL Server 2005 64-bit.
SCOM autoinstalls the right agents.
To add a new server, right-click in the Administration console and click Discovery Wizard. By default, it will scan your AD forest, but that’s not a good idea here since we’re just evaluating, so instead, click Advanced Discovery, click Next, click Browse For Computers, and click the Browse button to search the AD forest. It will ask for an admin account to install the agents, and click Other Account to use your own account if you’re an admin on that box (or a domain admin).
It then checks your servers to make sure they’re not already managed, and then asks you which management mode you’d like to use (agent or agentless). If you choose to use an agent, it will ask which account the agent should use – just use the system account.
SCOM2007 detects the operating system (Win2000 vs 2003 vs 2003 64-bit) and installs the appropriate agents.
Even better, it automatically detects supported applications like IIS, Exchange, and SQL Server, and installs the right agents for those as well. The whole process is very transparent – I didn’t even have to tell it which servers were running SQL, and which ones were not.
Bottom line: I can add a new server within 1 minute, which is way better than I’d expected. Note that after the agent is installed, it takes several minutes before health data is reported.
It’s easy to add and edit email notifications.
Adding new users takes less than a minute and requires no special skills. To add a new user, go into the console and click Administration, Notifications, Recipients, then right-click on Recipients and click New Notification Recipient.
Adding new subscriptions (alerts) is also easy. Right-click in the Administration console and click New Subscription Notification. A wizard steps you through all of the choices like which applications to monitor (Windows, IIS, SQL, etc), what levels of alerts (health, maintenance, performance, etc), and how often to send reminders.
It has a great monitoring dashboard.
The console can be installed on laptops easily, and can run over the VPN. It gives a great heads-up view of the servers.
It has in-depth application health reporting.
It has deep visibility into SQL Server health statistics, and even the default settings gave me useful information. For example, it alerted me when a database shrank itself, which is a major no-no for enterprise databases.
It has a maintenance mode.
I can put servers in maintenance mode before doing work on them. That way, when I’m rebooting a machine to fix something, no false alarm emails go out. I actively used this feature in ServersAlive at my last company, and I’ve asked for the same ability in UniCenter too. When a sysadmin group is really living by these alert emails, every reduced false alarm means a lot.
System Center Operations Manager 2007 Negatives
It still has annoying bugs.
For example, I can’t always edit my email subscription – when I right-click on it and click Properties, the entire System Center console crashes and quits. Another example is that personalized views aren’t always saved correctly, and the sort order is reset when you go from one screen to another. Maybe these bugs will be fixed in the final release, but again, maybe not.
You have to know your application’s vital stats.
Like UniCenter, SCOM2007 can send a dizzying array of notifications. For a successful implementation, the sysadmin of each box needs to understand what they’re monitoring. When I first told it to send me everything, despite only monitoring a handful of servers, I got about 150-200 emails per day – only about 10 of which were useful. I’m now in the process of whittling down the notifications.
The console requires horsepower.
It runs about the same speed as UniCenter’s tools on my P4 laptop, which is to say slow.
The dashboard is not web-accessible.
In order to manage SCOM2007, you have to install the console or remote desktop into a machine running the console. Long-term, I would prefer a web-accessible portal like UniCenter’s, and in a perfect world, it wouldn’t require Java either, so I could run it from my handheld. But now we’re just talking crazy.
What I Didn’t Evaluate
Server Uptime Reporting
It’s integrated with SQL 2005 Reporting Services and does some data warehousing as well. This was way outside of my scope of needs, though: I don’t anticipate long-term use of this as an uptime tracking tool or statistics analysis system. I’d rather get back to UniCenter for performance monitoring when its agents become available.
OM2007 has the ability to monitor systems without installing the OM2007 agent, and it can even audit event logs without an agent. For the purposes of this evaluation, I wanted to install the agents because I wanted in-depth reporting on the applications (SQL Server and IIS).
Operations Manager 2007 achieves the requirements so well and requires so little effort that I would recommend we deploy it for a short-term 64-bit monitoring solution.