Thursday, November 21, 2013

Ubuntu 3.11.0 kernel not as stable as hoped

As I mentioned in my previous post, Ubuntu recently released a 3.11.0 kernel for Precise.  Unfortunately, while it appeared stable during a long burn-in process with various pgbench workloads and scale sizes, it crashed shortly after putting the machine into service, so make sure you test thoroughly on your hardware. 

This particular hardware was the Dell Poweredge 720 that rackspace uses.

Guess we're going back to custom compiling kernels for a little while longer.

Anyone else have experiences good or bad with it?

I believe this is the same kernel that's shipping with 13.10 Saucy Salamander.

4 comments:

  1. I have an issue with Dell Poweredge 720 / Ubuntu 13.11 / kernel 3.11-04-generic (x64)
    If I set kernel.sched_autogroup_enabled=0, the kernel fails to boot and crashes.
    I've sent them a bug report: https://bugs.launchpad.net/ubuntu/+source/xubuntu-meta/+bug/1261744

    ReplyDelete
    Replies
    1. Make sure you have the very latest firmware. That solved our crashing problems on the r720.

      Delete
    2. Just to be sure, do you mean that your postgresql in now in production using ubuntu 3.11 stock kernel on 12.04LTS, thanks to a firmware upgrade on the r720, or that you still have crashes but can boot.

      Delete
  2. It is not in production, but it also has not crashed. We are still testing. It seems as though there is a performance regression in the PERC card compared with the r710.

    ReplyDelete