Data Repository sizing tool query

[Legacy User]

Hi
We are planning to upgrade Performance Management from 2.6 to 2.8 and I have run a sizing exercise, using the PM sizing tool, which indicates for the Data Repository 3 x Nodes, 8 x Cores and 64G Memory (8G per core).

However this is a small implementation, currently approx 200 devices, 16k items being monitored and the 2.6 system has been running fine for quite a few months on a single node, 2 x core, 16G memory system, so I am wondering if the upgrades to nodes, core and memory are actually needed.

 

Note all other systems, DA, DC and CAPC have the recommended specs.

 

The main thing I am wondering is why is there a need/recommendation for 3 nodes for the DR, is this for resilience?

 

And if I upgrade using the current systems is it easy to add in the extra DR nodes if I find they are needed?

 

Does anyone have any advice/opinions please?

 

John

[Troy Rondeau]

John,

 

The recommended specs listed from the sizing tool are there for both performance as well as resiliency.  The recommendations on performance are set to allow for expansion and tend to be nearer to the upper side of your configuration size (small, medium & large).  See below for a snippet straight from HP Vertica regarding this:

 

There are many reasons for adding one or more nodes to an installation of HP Vertica:

• Increase system performance. Add additional nodes due to a high query load or load latency or increase disk space without adding storage locations to existing nodes.

Note: The database response time depends on factors such as type and size of the application query, database design, data size and data types stored, available computational power, and network bandwidth. Adding nodes to a database cluster does not necessarily improve the system response time for every query, especially if the response time is already short, e.g., less then 10 seconds, or the response time is not hardware bound.

• Make the database K-safe (K-safety=1) or increase K-safety to 2.
• Swap a node for maintenance. Use a spare machine to temporarily take over the activities of an existing node that needs maintenance. The node that requires maintenance is known ahead of time so that when it is temporarily removed from service, the cluster is not vulnerable to additional node failures.
• Replace a node. Permanently add a node to remove obsolete or malfunctioning hardware.

 

As for adding another DR node to an existing environment, take a look at this link for the CA Performance Management Wiki:

 

Add a Node to the Data Repository Cluster - CA Performance Management - 2.8 - CA Technologies Documentation 

 

Please let me know if this answers your questions or if any more arise.

 

Troy

[Legacy User]

Hi
I have some additional questions so have marked this back as unanswered.

 

I guess I could rephrase my original question …

 

Do I NEED to upgrade the DR processing from 2 cores to 8 cores, and memory from 16G to 64G, simply because I am upgrading from 2.6 to 2.8? It seems excessive to me especially as the 2.6 system currently works fine with this config. Is the performance demand that much greater in 2.8?

Do you know if 2.8 will work with the current cpu/memory config or is there no point even trying?

Would it make sense to upgrade and see how things operate? Then if there are any issues I can subsequently upgrade the cores and memory.

 

I would appreciate if anyone in the community could give feedback on any similar environments and their experiences.

An additional question is that there is a note in the document you reference to add a node to a cluster which says “This process applies to adding a node to an existing cluster. To expand a single-server installation to a clustered installation, contact CA Support.”

Can you tell me exactly what is a single-server installation? I have a single DR node but it has the Cluster Management menu option and the option to Add Nodes to the Cluster etc. Is this a single-server install and do I need to contact support to add nodes?

 

Regards, John

[Anon Anon]

Hi John,

 

In the document where it references "a single-server installation" that is just referring to a single Vertica node (one Data Repository system).

 

As for the resources - Vertica is in fact very resource-intensive, especially with it's latest database versions. The published resource recommendations are from Vertica.

 

That being said - could you get away with fewer CPUs or less memory ? Perhaps, but the database performance will surely be impacted. I would venture that even though your current 2.6 system appears to be functioning fine, the query performance and all-around performance of the DR could be significantly improved if it met minimum specifications.

 

The flip-side is also that if you do move forward with your current configuration and start to experience database related issues, since the system wouldn't be meeting minimum recommendations one of the first actions in order to eliminate cause would be to bring the system up to spec.

 

I personally have many test systems which I have installed and used and those which were skimped on resources do perform much slow than those with adequate resources - with all systems otherwise being configured the same (same O.S. version, disk, number of monitored items etc.).

 

I hope this helps.

 

Regards,

Chris

[Legacy User]

Hi Chris

thanks for the reply. I will have a discussion with the client and recommend that they upgrade the CPU and memory on their system. If they aren't too keen to do that I will try the upgrade on their current system and review afterwards.

 

Just to be clear then, you are saying that their system is a single-server installation and as such I need to contact CA support to expand to a cluster installation. Is that correct?

 

Regards, John

[Anon Anon]

Hi John,

 

Correct. They should have a support ticket created so that we can advise and assist in adding nodes to the Data Repository. We have an implementation doc for that. 

 

Regards,

Chris

[Legacy User]

Hi Chris

The customer held off on the upgrade to PM 2.8 and is now looking at going to 3.0. So I reran the sizing and there appears now to be an option to run the DR on a single node, see screenshot.

 

Can you confirm this is the case please?

 

Regards, John

[Michael Poller]

Vertica has always been capable of running in single node form.

 

It isn't really how the HP Vertica DB product is designed to work. The intended configuration is a multi-node cluster, the minimum being a three node cluster.

[Legacy User]

So what is the "Data Repository - Single Option" for?

 

Is it a supported configuration?

 

Regards, John

[Bob_Keville]

Hi John - we think we have answered this question for you - please let know if that's incorrect

[Legacy User]

Hi

I have some additonal questions so will be marking this back to unanswered.  I would appreciate if in the future you left it to me to mark my questions as answered when I feel they have been answered.

 

Regards, John