What is the best ditributed DB deployment for Orion modules

Hello,

We are having performance issues with our Orion SQL DB, causing all sorts of problems ranging from gaps in the graphs to IPAM not scanning properly, to Orion pages not loading and giving "unexpected error has occurred" messages instead.

We have raised many tickets with support, and most of the feedback was that our DB is too big, SQL jobs are not finishing, etc .... they would sometimes give us some queries to run which help for some time, only to come the next morning and find that some pages are not opening again.

To stop complaining and actually get to my question, i was thinking of some kind of a distributed DB architecture, in an attempt to ease the pressure off our main (and only) DB.

We currently have 2 polling engines and 1 SQL server.

Has anyone attempted such a deployment? Maybe tries to install a separate SQL server and use it for NTA and/or NCM ?

Thanks,

Find more posts tagged with

distributed_db

Performance

graph_gaps

sql

distributed_sql

gaps

unexpected_error_occurred

orion_database

Accepted answers

All comments

savell

How many elements are you polling and how big is your database BK?

We have 4 pollers (about 20,000 elements) with a database size of approx 40Gig - running NPM, APM, Netflow (relatively small) and IPSLA.

Our NCM implemenation is intergrated, but the database lives on another server.

If you have enough server memory and good hard drive speed (a resonable number of 15k drives) then I wouldn't have thought you would have any issues with two pollers (perhaps you have huge amounts of NTA data?).

Both of our NCM and NPM databases each live on a couple of mirrored Fusion IO SSD drives on our standard DL380 servers - I/O speed issues are something I don't worry about these days.

Dave.

bkattan

Thank you for your reply savell.

Our DB server has 4 10K RPM drives in a RAID0+1.

Our database size is around 270 GB.

The following is our server details:

Main Orion Server Details

Orion

Module Name	Orion Core
Version	2011.2.2
Service Pack	None
Nodes currently monitored	2681
Total nodes in license	unlimited
Volumes currently monitored	202
Total volumes in license	unlimited

NCM

Product Name	Orion Network Configuration Manager
Version	7.0.2
Service Pack	None
License	Production
Allowed number of nodes	unlimited
Current number of nodes	2291

SAM

Product Name	Server & Application Monitor
Version	5.0.0
Service Pack	None
License	Production
Allowed Number of Component Monitors	1500
Total Number of Component Monitors	135
Licensed Component Monitors	135
Unlicensed Component Monitors	0
Available Component Monitors	1365

NPM

License	Production
Product Name	Network Performance Monitor
Version	10.2.2
Service Pack	None
Current number of interfaces	6935
Allowed number of interfaces	unlimited

IPAM

Product Name	IP Address Manager
Version	3.0
Service Pack	None
License	Production
Allowed number of used IP Addresses	unlimited
Current number of used IP Addresses	33511

IPSLAMGR

License	Production
Module Name	IP SLA Manager
Version	3.5.1
Service Pack	None

NTA

License	Production
Module Name	NetFlow Traffic Analyzer
Version	3.9.0
Service Pack	None

IVIM

Module Name	Integrated Virtual Infrastructure Monitor
Version	1.2.0
Service Pack	None

And the following is our retention:

Deltona

FYI, RAID 0+1 is NOT the same as RAID 1+0 (RAID 10)

How many databases are hosted on the SQL server and what are their functions?

Have you checked to see if the disks are functioning properly? Any errors?

bkattan

Thanks Deltona,

Can you help me understand your comment on RAID ? RAID 0+1 is the one recommended by support.

mavturner

bkattan, do you regularly use all of this historic information? It looks like you have increased the defaults of all of the historical retention settings. If you do not use this data regularly then I suggest putting them back to default. If you are required to keep it for compliance requirements but don't actually access the data regularly then you might consider storing that historical data on a separate server (data warehousing).

bkattan

Very good point mavturner.

I do need the stats for compliance actually, but do not access them regularly.

So how do you advise I proceed with data warehousing?

Should i set the retention setting back to default and have some automated process regularly move the data to some other database?

If so what is the method you recommend?

And if i do that, can i still see those stats in Orion?

mcbridea

We recommend RAID 10. A lot of RAID controllers have performance problems with RAID 01. Here are a couple of Tech Refs that address the SQL server and the Orion database.

http://www.solarwinds.com/documentation/Orion/docs/OriondbBestPractices.pdf

http://www.solarwinds.com/documentation/Orion/docs/OrionPerformance.pdf

Andy