CA PM - Min. Number of Data Repository Nodes- 3?

[Phani Devulapalli]

Hello All,

 

Is it mandatory to have at least 3 nodes part of the Data repository ? I see the sizing tool does not allow anything less than 3 . I understand there are certain advantages of a 3 node cluster instead of a single node DR but in certain scenarios the resource requirements (CPU, Memory, Disk) for a 3 node cluster are quite enormous and if we decide to go with a single node would there be any issues ?

 

Are there any customers who are running CA PM in a single node or does everyone choose to go with a multi node cluster for DR?

 

Please let me know your thoughts 

[Lutz_Holzbecher]

Hi Phani,

all the time we have had customers running with single node DR/Vertica system, especially in smaller scale environments as described here Install a Low-Scale System - CA Performance Management - 3.5 - CA Technologies Documentation  

The sizer reflects this option as well when scaling e.g. a system with less than 150.000 polled items ( e.g. 3.500 devices).

For larger environments we recommend 3 node (or even more) clusters, which not only tolerate single node failures but also provide distribution of query load.

Hth,

Lutz.

[Phani Devulapalli]

Hi Lutz,

 

Thanks for your comments, the number we are looking for is way above 3500 device limit so may be we should consider a 3 node cluster . I agree there are surely advantages of having a 3 node cluster for a large environment, but resource requirements provided by the sizing tool looks enormous when considering the CPU, Memory and Disk ( data+backup) compared to what we have been using for eHealth. 

Anyways we will consider our resource availability and make a choice 

 

Thanks, Phani

[Troy Rondeau]

Hi Phani,

 

Just wanted to see if you had any other questions regarding the Vertica configuration after Lutz's response.  As for the sizing in comparison to eHealth there is definitely a substantial difference when considering the DR component in particular.  It is this sort of requirement that allows the k-safe functionality and adds the extra redundancy that keeps you up and running if a node goes down.

 

Troy