Freezing user interfaces

A number of machines have, after the upgrade this summer, seen the user interface freeze or become unresponsive. Today the sysadmin staff has pushed a change to all machines with the affected hardware in the hopes of resolving this issue.

The problem is rather elusive from a systems point of view. Machines that use Nvidia cards that have Tesla-based chipsets experienced intermittent issues with clients (in this case, applications) disconnecting from the X server itself. This would lose mouse and keyboard to most applications, except a select few. The drivers affected are the latest open source nouveau as shipped by RedHat Enterprise Linux, and the appropriate Nvidia proprietary drivers (the 340xx-branch) cause kernel panics. This leaves us without a functioning driver for the latest kernel for these cards. In an attempt to bypass this issue, we have rolled back the kernel and the graphics driver on the machines in question to the latest RHEL7.3 kernel. From a user perspective, the change should improve stability and not have any effect on the use of the machine. We apologise for the issues this issue has caused.

Published Aug. 17, 2017 9:24 AM - Last modified Aug. 17, 2017 9:24 AM