Chat - HA failover is not working
Hi,
We have 2 nodes for chat application and lets say C1 and C2.
If a chat request has gone to C1 and is currently in-progress and due to some issue C1 has crashed.
- Will the in-progress chats on C1 will be routed to C2 and continue with out any glitch?
- If chat is crashed, will customer know that chat has crashed and displayed some error message
FYI- We are running chat servers with forever.
***Updated by moderator: Lochan to add SR to post***
Accepted Solution
Pegasystems Inc.
US
Upon reviewing the associated SR, SA-55635 was generated as the resolution.
If you have the same question, please check out "Chat client do not reconnect at node failure" to resolve.
Pegasystems Inc.
IN
I am assuming that you have a load balancer before C1 and C2, so if for any reason C1 crashed the customer should connect to C2.
Further, if the customer is connected back to C2, he will get appropriate disconnect message
Accenture
AU
Hi,
We have load balancer infront of app servers and when C1 is crashing, all the chat requests that are on C1 are getting disconnected and no message is displayed to customer. FYI - We are using NetScaler and LB's are enabled for sticky sessions.
Is there recommended settings for NetScaler's LB?
Pegasystems Inc.
AU
From install guide:
https://docs-previous.pega.com/premises-chat-server-installation , page 5 , Recommended setup:
It is recommended that the chat node.js servers sit behind a reverse proxy load balancer.
− It is required that the load balancer support sticky session routing (for example, based on IP address) and WebSocket proxying. An ideal choice would be NGINX.
From the NGINX website:
High-performance load balancing , refer to Load-Balancing Methods.
On the point of NetScaler, As discussed this is not tested inhouse, but we will check with the team on any knowledge on NetScaler, but I would strongly suggest that you discuss with your team on how to setup NetScaler to do routing based on IP address.
Updated: 15 Mar 2018 2:07 EDT
Accenture
AU
We have had a look at LB configurations for NetScaler and our network team has confirmed all the configurations are setup properly. As you already know, I've raised an SR for tracking this. SR-C19868
Accenture
AU
I have few more queries regarding how chat works on LB
- Can you explain how would chat work in front of LB and how does routing from one node to another node happens when a node is down?
- If LB just routes the requests, how would LB be able to route a session from dead process to alive process?
- Is the above a feature of product capability. If so, is it available in Chat version 7.22.19?
Thanks!
Accepted Solution
Pegasystems Inc.
US
Upon reviewing the associated SR, SA-55635 was generated as the resolution.
If you have the same question, please check out "Chat client do not reconnect at node failure" to resolve.