HOW TO: Configure Infiniband Active-Active Network with Failover/Failback Support Using RDS "active-bonding" Feature in Oracle Linux
(Doc ID 2135112.1)
Last updated on FEBRUARY 03, 2019
Linux OS - Version Oracle Linux 6.0 and later Oracle Cloud Infrastructure - Version N/A and later Linux x86-64
This document will provide guidance on how to setup an Infiniband Active-Active Network with Failover/Failback Support using RDS bonding feature in Oracle Linux using UEK Kernel
Recent kernel version v2.6.39-400.278.1 which contains all bugfixes which can affect RDS active-active operation is recommended. Please refer to section "5. Recommended kernel and HCAs for using RDS Active-Active configurations with Failover/Failback"
Import to consider:
The failover (and failback) happens to (and from) ports within the same HCA.
Ports in the HCA should ideally be wired through different switches to allow for the use case of a switch reboot such as during switch firmware update without disrupting operations.
Failover/Failback implemented in kernel as part of 'rds_rdma' module
The feature often uses the name 'active-bonding' but is performed by migrating IP addresses and not to be confused with any Linux bonding driver modes. It DOES NOT use the Linux kernel bonding driver.
Please do read "Caveats and Limitations" section later carefully when using this feature.
This document will not provide information about how RDS and infiniband works. The purpose of the document is to help the user with its configuration provinding examples of configuration and results of testing.
There will be a "caveat and limitation" section at the end of the document that we recommend the Reader to review.
To view full details, sign in with your My Oracle Support account.
Don't have a My Oracle Support account? Click to get started!