Question
Cognizant Technology Solutions
US
Last activity: 3 Apr 2024 0:33 EDT
How does a non web user node recognise or communicate with a Pega web user node?
I understand that a LB is set up for the web users to direct the traffic to web user nodes. But I would like to understand how a web user node recognises that it needs to run a job scheduler on a background processing node ?
In a mutli node environment, how does each node recognises each other ? Is it via the "Data-Admin-Nodes" instances ? I'm assuming Pega engine installed on each node behind the scenes checks this database to route the task internally. Can someone correct my understanding ?
Maantic Inc
IN
@KVPVishnu The configuration you are searching for you can check the setenv.sh file in EC2 or Unix shell server box. based on which the Admin studio identifies nodes as webuser or nodes as stream or background. And every node communicates with other nodes using Pega Pulse or Pega Hearbeat keeping in sync.
Cognizant Technology Solutions
US
@Soham_Chanda1107 I understand that we can classify them in that set env using DNodeType variable but my question is,
Although we install Pega platform in 3 nodes and classify them as - Webuser, Background & Stream respectively, how would each of those nodes know that they are part of Pega cluster ? Like there should be some linking of those IP addresses to determine that they are part of 1 cluster. Isn't it ?
NCS Pte. Ltd
SG
Hi @KVPVishnu: Node classification helps you to define the purpose of the nodes. You can tag your QP and JS to appropriate classified nodes.
Please see details here.
In terms of internode communication, PEHA uses hazelcast. Please see details here.
Please also see this regarding QP functionality.
Thanks.
Cognizant Technology Solutions
US
@ArulDevan So, in simple words, when we schedule a job scheduler on a background node, internally among the cluster of nodes, Hazelcast is something that Pega uses to let the background node know that, there is a job scheduler for it to execute ?
NCS Pte. Ltd
SG
Hi @KVPVishnu: Nope. Hazelcast is used to maintain the cluster state. When comes to Job scheduler, it is the node classification which helps to determine where the Job scheduler to run. For the Queue processors again it is the node classification which identifies where the job scheduler to run but the items will be queued to kafka.
Thanks.
Cognizant Technology Solutions
US
@ArulDevan exactly my doubt here. Let's say we classify three nodes as Web, background & Stream. Now, I'm on web-user and scheduled a job scheduler to run on bg node. Now, by classifying the nodes, we are only mentioning that it should be run on Bg node. But, however, we're not mentioning any further details of that background node on the web-user node server configuration details.
So, how would the Web-user system recognize the background node internally without specifying the Bg node's details (like ip address, port etc) explicitly ? is it something that pega engine code takes care of when all those 3 nodes are part of cluster (when connected to same DB, I'm calling it a cluster) ?
NCS Pte. Ltd
SG
Hi @KVPVishnu:
Note sure on what is the use case for a web node to recognize Background processing node. When you use job scheduler, you will define in the job scheduler to say what node type it should run. When the job scheduler wakes up in the corresponding node, it will process based on the logic configured. Please see more details about job scheduler here.
If you are looking for Queue processors, it is managed using the stream nodes. Please see detailed explanation here.
If you are looking from a development perspective where you create a new Job scheduler, its all connected to the same DB and your Rules will be saved in same DB. But the Job scheduler will run based on node classification configured in the rule.
Thanks.