Re: How do you manage Databases?

I don't see any responses to this. Maybe I just haven't picked them up. But I'll throw out some of my patented brutally-honest, sure-to-offend, but practically-based experience:

ekrieter_at_mcs.net wrote:

   [...]
> I manage the network infrastructure
>via HP Openview running on a Sun Solaris/Ultra Enterprise 2. I am under
>the impression that I will be able to reasonably be able to monitor
>things outside of the databases. By things I mean disk capacity, CPU
>utilization, memory usage and others. Are you monitoring these things
>also? What suggestions might anyone have?

You can go with a proprietary agent (BMC, IBM, CA) or a private SNMP agent (usually vendor-specific, but the freebie UCDavis host agent has some interesting capabilities for monitoring processes...), or you can stick with an agent that support the standard HOST RESOURCES MIB (Empire Technologies is probably the best). Any of these can give you "host stats" for CPU, memory, and disk resources. Most of the proprietary agents have "agent-side thresholds" which can be problematic. We generally recommend manager-side thresholding with a vendor-independant manager side (Seagate NerveCenter for real-time alerts, Concord Server Health for trending).

>
>What about internally to the databases? Here is where I'm less informed.
>Could I do anything to tables from Openview without adding additional
>products? I am aware HP has Smart plugins for Oracle? Does anybody
>like these addins? For some reason, they may require IT/Operations. Can
>we get by with just HPOV NNM? I've heard IT/O is expensive(very),
>anybody know a ball park price?

Monitoring databases is the real bitch, yes. Nobody's gotten anywhere much accept on table stuff (and certainly no control, though simple correction might be feasible in limited circumstances). You have to get some rather pricey special-purpose agents (BMC has one of the widest bases, though their sales channel is rather gummed up). The most likely approach is simply to push the database server vendor to enhance their own SNMP accessability. If you want to truly monitor the server processes and the table stuff, if its simply expensive, you're getting off easy!

The real problem is to do this stuff adequately, you need two or three agents to monitor the host/hardware platform, the operating system/server processes, and the database server/table mechanics. Until industry-wide extensible agent technologies are implemented (give it 5 years, at least, would be my bet, though you might start to play in the next year or two), this can be incredibly gory. OV can recognize alternate ports, but it can't use more than one port per IP address. All other software is likewise limited, as far as I know.

>
>How about the Patrol agents? Do they integrate at all with Openview?

As I mentioned, BMC Patrol is probably one of the cleanest and most capable effort in this vein, but be prepared for never quite getting what you want, and some healthy development time, to boot.

>
>And what about HP's own snmp Extensible agents? What can they do in
>regards to database monitoring?

Nothing, AFAIK, but then, I'm unaware their agent is extensible.

>
>I look forward to hearing about your experiences! Thanks!

Sorry, the caveat "there's not a lot of experience here, but that's why this is hard" needs to be reiterated. I'd certainly love to hear of some solid solutions from the database groups, but there doesn't seem to be anything here in the management groups. BTW, could somebody drop me an email if I'm missing non-cross-posted discussions, maybe?

--

T. Max Devlin
Hi-TECH Connections/Eltrax Systems
*****************************************************
 -   Opinions expressed are my own.
       Anyone else may use them only in
        accordance with licensing agreements.   -