kernel-packages team mailing list archive

[Dan Streetman <dan.streetman@xxxxxxxxxxxxx>]

>  I won't pretend to know how numactl interleaves the memory across the nodes,
> but I can't help but think high memory usage on these nodes combined with
> forced interleaving might be why we hit this issue?

The numactl interleaving just causes memory to be allocated from all
nodes on a round-robin basis, I don't think that would cause this, other
than mongod simply using a whole lot of memory.

> After weeks of stress testing with your custom kernel, I have yet to
hit this issue again

so, the custom kernel actually bypasses the BUG() call, and logs debug
instead - have you checked your logs to see if there are any relevant
messages?  You would see output like:

page_zone(start_page) !=page_zone(end_page)

and more debug following it; you can search/grep the logs for
"move_freepages".

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1497428

Title:
  kernel BUG at /build/buildd/linux-3.13.0/mm/page_alloc.c:968

Status in linux package in Ubuntu:
  In Progress
Status in linux source package in Trusty:
  In Progress

Bug description:
  The kernel triggers a BUG when it finds it is in move_freepages() but
  the start and end pfns for the move are in different zones.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1497428/+subscriptions