Why do frames get dropped with ICMP destination unreachable fragmentation needed after upgrading to RHEL 6.5?

Solution Unverified - Updated -

Issue

  • Ever since upgrading all of our hypervisors to the latest 6.5, and also upgraded our RHEVM environment to 3.2, we are getting "ICMP Fragmentation needed and DF set" for TCP data packets greater than 1479. This is causing a lot of data packet re-transmissions and a slowness of about 3.4 seconds per transaction between out applications. The source is a Solaris 10 server and the destination is a RHEL 6.2 Virtual Machine that sits behind a Red Hat Piranha load balancer.
  • We see that RHEL 6.5 is re-assembling the fragmented packets and trying to send a bigger packet where it is getting dropped by receiver due to limitation of the MTU size.
  • We see that RHEL 6.4 is not re-assembling the packets and there is no issue as it is sending only fragmenting packets.

Environment

  • Red Hat Linux Enterprise 6.5 and later
    • kernel-2.6.32-431.el6 series and later
  • Also affects products based off RHEL 6.5, such as the RHEV Hypervisor.
  • Forwarding of IPv4 or IPv6 traffic (affects both regular IP routing as well as bridging)

Subscriber exclusive content

A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.

Current Customers and Partners

Log in for full access

Log In

New to Red Hat?

Learn more about Red Hat subscriptions

Using a Red Hat product through a public cloud?

How to access this content