Unable to start expected number of vServers on an Exalogic Virtual Compute Node
(Doc ID 1612868.1)
Last updated on FEBRUARY 25, 2019
Applies to:Oracle Exalogic Elastic Cloud Software - Version 184.108.40.206.0 and later
Oracle Virtual Server(x86-64)
You have a new installation of Exalogic Elastic Clooud Software 220.127.116.11.0 deployed in a virtual configuration and may have successfully created a number of vServers that have previously started successfully on one or more compute nodes. However, you are now experiencing a situation where, despite the availability of the required vCPU and Memory resource, the creation or start up of a vServer fails to successfully initialize the network environment and IPoIB or EoIB networks are not accessible.
When impacted by this issue you may be able to start exactly the same vServer on a different compute node where fewer vServers are running, which proves there is nothing wrong with the vServer configuration or the network infrastructure on Exalogic switches that has been created to support the networks associated with the vServer. However, you will find that each compute node will not successfully start of as many vServers as should be expected given the RAM and vCPU resources available on affected nodes and that the number of vServers that can be successfully started can vary significantly: many times up to 4 vServers can be started before the next vServer to start encounters the problem whilst, at other times, fewer or no vServers can be started without encountering the problem.
The issue is most likely to be encountered on X3-2 systems (which by default have 256 Gb RAM) or X2-2 systems where the memory has been upgraded beyond the default level of 96Gb installed from the factory. The amount of memory present on a given compute node can be determined by executing "xm info" as illustrated below:
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