Known Issues

Each known change request (CR) is listed by its synopsis and its CR number. Each entry describes the issue and provides a workaround, if one is available. Check the latest status of CRs at:

http://support.oracle.com

Links Do Not Work Correctly After Running Oracle VTS and Internal Loopback Test Fails (CR 7147018 and CR 7008767)

Performing an internal network loopback test with Oracle SunVTS on Sun Quad Port PCIe 2.0 Gigabit Ethernet Networking Cards might result in a test failure and improperly working links.

Workaround: Avoid using Oracle SunVTS network loopback test. If you ran the test and the links are not working properly, try the following solutions, depending on the operating system.

For the Oracle Solaris 10 OS, perform these steps:

1. Reboot the system.

2. Power cycle the system.

3. Download, and apply the latest patches.

   If downloading the latest patch does not solve the issue, contact your Oracle service provider.

For the Oracle Solaris 11 OS, update the system with the latest service patches.

Network Latency Under Oracle OpenSolaris Is Three Times Slower Than Solaris 10 5/09 (CR 6933376)

Starting with Solaris 10 10/09, you could see network latency spikes. This issue is related to the Intel power management feature and deep C-state. When a CPU is idle, it can either halt until it is woken up, or the CPU can spin. MWAIT and HLT can halt a CPU, but they have different wakeup mechanisms. Deeper C-states are greater power-saving halt states, but they can also cause longer wake-up times.

Workaround: Add these entries in the /etc/system file then reboot both systems.

...
set idle_cpu_prefer_mwait=0
set idle_cpu_no_deep_c=1
...

When only idle_cpu_no_deep_c=1 is set, the system only uses the C1 halt state. Thus, you must set both states. This works with either HLT or MWAIT.