Hi Ed,
I'm still working on getting the RHEL5.3 install, so I have not yet been able
to reproduce your result. I expect the system should be ready tomorrow (10/6).
While I have been waiting, I have performed tests of bonding with (almost
certainly) more recent versions of the bonding and intel drivers, and at least
I do not see the problem you report on these.
The bonding module has a number of options available for how it determines
whether a link is available to select as active. The preferred method is to use
the subtended device's netdev link state, which reliably works for me in my
tests. In fact it is the default method at bonding version 3.5.0, which is what
I am currently using. I am not sure what's in your distribution, but suspect it
may be using the older link query method (user_carrier=0). Here are my settings.
modprobe bonding mode=1 use_carrier=1 miimon=100
The reason I suspect use_carrier may be relevant is because without it, bonding
will query for available link using either SIOCGMIIREG or SIOCETHTOOL methods,
which will return the true layer 2 link state, not quite the same as whether
the interface is ready to TX/RX traffic. Further, there are existing problems
with these methods, which I am investigating on another track.
If use_carrier is an available module load parameter for your installation,
could you please try it and see if bonding now properly switches adapters.
Thanks
Dave
-----Original Message-----
From: Ed 177215 [mailto:[email protected]]
Sent: Friday, October 02, 2009 6:21 AM
To: [email protected]
Subject: [E1000-devel] Teaming unsuccessful - e1000e-1.0.2.5 (2.6.18-128.el5 &
2.6.18-164.el5)
Hello e1000 Developers,
I have a RHEL 5.3 (2.6.18-128.el5, driver e1000e-1.0.2.5 built) machine,
with two network adapters, which I'm trying to get to Team/Bond - an (i)
Intel PRO/1000 PT Server single-socket, and (ii) Intel PRO/1000 GT Desktop
single-socket. Individually, the cards work fine. Also, their Team (mode 1)
setup is fine, and logs show Teaming to be working.
However, Teaming does not actually work, and the issue appears to be very
close, indeed the same, as that described in
https://bugzilla.redhat.com/show_bug.cgi?id=483034 (Clsoed Duplicate). In
short, when a fault occurs on the active Team member (eth0), the switch
occurs, but the machine is no longer pingable/accessible, until eth0
eventually becomes the active member. Again, the bonding/teaming log shows
the correct bonding status on the cards and bond/team at all times.
I upgraded the RHEL box from 2.6.18-128.el5 to 2.6.18-164.el5 (so, RHEL
5.4), but the issue still occured.
So, I tried rebuilding the e1000e-1.0.2.5 src (previously built for
2.6.18-128.el5) to match the 2.6.18-164.el5 kernel. However, this resulted
in a build error -
".../e1000e-1.0.2.5/src/kcompat.h:1505: error: redefinition of 'struct
napi_struct'
Googleing this build error, I found that it's been commented on in this
mailing list recently (
http://www.mail-archive.com/[email protected]/msg01883.html),
though for a different build.
If you'd care to comment on the above path to get Teaming/Bonding working,
this would be much appreciated.
Regards,
Ed
------------------------------------------------------------------------------
Come build with us! The BlackBerry® Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay
ahead of the curve. Join us from November 9-12, 2009. Register now!
http://p.sf.net/sfu/devconf
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
