Red Hat Bugzilla – Bug 124648
tg3 not recovering from disconnects
Last modified: 2007-11-30 17:07:02 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.6)
Description of problem:
IBM xSeries 445 with two NetXtreme gig fiber cards using "tg3" driver
(with and without "bonding" driver). Networking becomes unstable when
connectivity restored (after some interruption). When bonded, the
switch will go "amber" on one of the NICs (seems to alternate). We can
get decent network connection if we change config from 2 bonded NICs
to 1 NIC (sometimes eth2, other times eth3, almost always
alternating...coincides with "amber")
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Bond 2 NICs
2. Disconnect cable (switch goes amber)
3. Reconnect cable (switch stays amber)
Actual Results: SSH connection becomes choppy
Expected Results: Degradation should have been minimal/unnoticable
I've just finished looking over bonding.txt and I think I might see
the problem. I don't believe our switches support modes 0,2,3,4 and I
must be using the default (mode 0). I'd like to use "balance-alb"
(presumbably that would be the best for an Oracle database server),
but I don't know that the tg3 drivers supports this mode on our NIC
(IBM NetXtreme 1000 SX Fiber a/k/a BCM5703). How can I find out what
modes are supported, and what is a recommended value of miimon for
Okay, I've tried the following in /etc/modules.conf:
options bonding miimon=100 mode=balance-tlb
options bonding miimon=100 mode=balance-alb
options bonding miimon=100 mode=active-backup
and all have produced the same error condition. FYI, these cards are
linked to a Nortel Passport 8010 which, according to our network guy,
has no support for link aggregation (which presumably prevents me
from using any mode but these three).
Out of curiosity, i rebooted the computer w/o bonding, so that I
could have both switch lights stay green. I then changed to a bonding
restarted networking and both switch lights stayed green (one went
amber for an instant, but back to green)
I've had this condition before, but invariably, a network
interruption would cause something to flake and my ssh connections
would be choppy again. This is the first time I have it now that I'm
trying "mode=balance-tlb". However, when I disconnect the cable and
the switch light goes amber, it still stays amber when I reconnect. I
ran an SSH session which started out fine, but suddenly disconnected.
When I tried re-establishing, I couldn't even connect and a ping-to-
switch I started on the server before disconnecting has suddenly
begun losing packets.
This bug is filed against RHEL 3, which is in maintenance phase.
During the maintenance phase, only security errata and select mission
critical bug fixes will be released for enterprise products. Since
this bug does not meet that criteria, it is now being closed.
For more information of the RHEL errata support policy, please visit:
If you feel this bug is indeed mission critical, please contact your
support representative. You may be asked to provide detailed
information on how this bug is affecting you.