Red Hat Bugzilla – Bug 1464938
GlusterFS plugin should read 'endpoints' from a global namespace
Last modified: 2017-10-21 13:10:25 EDT
3. What is the nature and description of the request?
The current GlusterFS plugin for OpenShift requires a service and endpoint within every project which may potentially connect to a GlusterFS PV. Instead we want glusterfs plugin to read endpoints from a gloabal namespace
4. Why does the customer need this? (List the business requirements here)
Based on the current scenario we need to have service and endpoint within each project. This causes problems when you start to scale the cluster as you can have, for example, an additional 1000 Services when you have two clusters and 500 namespaces, this is actually tripling the total number of services in our cluster)
Each of those services then requires extra entries in the iptables NAT tables. And since the iptables implementation in OSCP 3.5 deletes and recreates the whole entire iptables ruleset, this blocks the node processes and results in excessive load on every node in the cluster, and thus in problems scheduling pods.
We request that the GlusterFS plugin should be able to read data from a global namespace, prior to checking the 'local' namespace so that we don't need to maintain a ridiculous number of IDENTICAL services and endpoints across every namespace.
5. How would the customer like to achieve this? (List the functional requirements here)
Glusterfs plugin should read from global namespace depending on a parameter when enabled in say master-config.
6. For each functional requirement listed in question 5, specify how Red Hat
and the customer can test to confirm the requirement is successfully implemented.