Questions Regarding Scaling WCS Remote Satellite Server (RSS) For Heavy Load Scenarios
(Doc ID 2319287.1)
Last updated on OCTOBER 29, 2021
Applies to:Oracle WebCenter Sites - Version 22.214.171.124.0 and later
Information in this document applies to any platform.
Question 1 - Connection pooling
It appears that for a single page request, 16 calls are being made to the sites server from RSS, via LB/OHS, each one of them creating a new connection. The connections are never re-used for the same session or for different sessions. This is highly inefficient.
Question 2 - Connection closure logic
a) It appears that 2 connections are left in an established state when the back-end server (i.e OHS/LB) has keep-alive ON and that is persisted for the keep-alive time honored by the server. This happens for every new request (since it doesn’t re-use connections). This leads to holding up of socket connections when there are too many requests. Also, this logic seems erratic since only 2 out of 16 connections are kept-alive in established mode for the keep alive duration.
b) When internal LB is used, many connections are seen left in time-wait mode (60 seconds) on the RSS side. However, this behavior is not observed when OHS is used. The logic from RSS side needs to be consistent. RSS needs to send keep-alive off explicitly if it is not going to pre-close connections to avoid confusing the back-end servers.
Question 3 - Clustering support
Clustering support for the back-end servers (i.e., support for a number of LB’s or OHS’s so as to provide availability and scalability).
Question 4 - Configure polling
Ability to mark a back-end server as “un-available” and poll at a configured frequency to check its availability. This would avoid unwanted retries.
Question 5 - Clustering for High Availability (HA)
How we can achieve clustering on the RSS side for HA.
To view full details, sign in with your My Oracle Support account.
Don't have a My Oracle Support account? Click to get started!
In this Document