Re: [libteam]teamd_init() failed
by Wenhua Zhang
>> >>
>> >> Hi Flavio,
>> >> As we know, if the bonding mode set to be 4(Dynamic link aggregation),
>> >> it requires special switch configuration.
>> >
>> > Not just that mode.
>> >
>> >> Do you mean that if we use libteam with teamd, we have to do some
>> >> special switch configurations? That is what we plan to avoid.
>> >
>> > Depending on the mode you choose, yes. Active-backup mode, for
>> > instance, doesn't require switch config.
>> >
>> >> Because for some switch(eg, H3C S5120-52C- EI ), the number of
>> >> aggregate ports has a upper limit, and there will be some ports can't
>> >> be aggregated in the switch.
>> >
>> > That's indeed a problem. The only way to work around this
>> > but keeping the same mode is to use different switch to
>> > accomplish unique paths to the destination.
>> >
>> > eth0 ---> switch 0 ----> eth0
>> > System A eth1 ---> switch 1 ----> eth1 System B
>> > eth2 ---> switch 2 ----> eth2
>> >
>> > The above setup doesn't care that system A's eth{0,1,2}
>> > have the same MAC address because they are in isolated
>> > physical paths. I never tried that myself though.
>> >
Hi Flavio,
The net topo we have as below:
The eth1-4 connect to the same one switch, and for teamd with loadbalance
mode, p1/2 will have the same MAC(changed by libteam), and p3/4 will
have the same MAC.
+------------------------------+
| switch |
+------------------------------+
p1 p2 p3 p4
| | | |
| | | |
eth1 eth2 eth1 eth2
| | | |
+-----+ +------+
A | | B
+------+ +-------+
As we know, for bonding driver, the 802.3ad mode requires that the
switch have the appropriate ports configured as an 802.3ad
aggregation, what about libteam with loadbalance mode?
Can you show me some config to do for the switch with loadbalance
mode? And what about lacp mode?
Thanks,
Best Wishes,
Wenhua
11 years, 6 months
Re: [libteam]teamd_init() failed
by Flavio Leitner
[please, don't remove the libteam list from CC, others might
find this discussion useful]
On Fri, 1 Jun 2012 00:26:15 +0800
Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> Hi Flavio,
> Can you show me your config file and the environment you use?
Sure, I have two configs currently working:
# cat team0-activebackup.cfg
{
"device": "team0",
"hwaddr": "00:1C:C0:A3:7F:16",
"runner": { "name": "activebackup" },
"link_watch": { "name": "ethtool" },
"ports": {
"em1": { "prio": 100 },
"p11p1": { "prio": 100 }
}
}
# cat team0-lacp.cfg
{
"device": "team0",
"hwaddr": "00:1C:C0:A3:7F:16",
"runner": {
"name": "lacp",
"active": 1,
"fast_rate": 1,
"tx_hash": ["eth", "ipv4", "ipv6"]
},
"link_watch": { "name": "ethtool" },
"ports": {
"em1": { "prio": 100 },
"p11p1": { "prio": 100 }
}
}
# uname -r
3.4.0-01622-g31a6710
# lsmod | grep team
team_mode_loadbalance 1819 1
team_mode_activebackup 1690 0
team 18475 2 team_mode_loadbalance,team_mode_activebackup
> I am confused about how to test the libteam, and I hope you can give
> me some advice.
> We plain to use libteam instead of bonding, and before that we need do
> some tests.
>
Certainly. I am testing 'teamd' here as well, so I appreciate any
feedback you can give to us.
thanks,
fbl
> Thanks,
> Best Wishes.
>
>
> 2012/6/1 Flavio Leitner <fbl(a)redhat.com>:
> > On Thu, 31 May 2012 23:23:42 +0800
> > Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> >
> >> Hi Flavio,
> >> Thanks for your help.
> >> I have build the latest net-next kernel and try again.
> >> There is still some problem, and I will try some other method to fix it.
> >>
> >
> > I did those steps yesterday using net-next and I am now running
> > active-backup on two different machines.
> >
> > I had to clean the 'configure' caches from libnl and libteam as well
> > in order to get updated binaries.
> >
> > fbl
> >
> >>
> >>
> >> 2012/5/31 Flavio Leitner <fbl(a)redhat.com>:
> >> > On Thu, 31 May 2012 11:36:43 +0800
> >> > Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> >> >
> >> >> Hi all,
> >> >> When I try to test the libteam as Example in
> >> >> https://fedorahosted.org/libteam/wiki/InfrastructureSpecification, I
> >> >> got errors as blew:
> >> >> Failed to create team device.
> >> >> teamd_init() failed.
> >> >> Failed to start daemon: Operation not supported
> >> >>
> >> >> The command is :teamd -f teamd2.conf -D
> >> >> The conf is :
> >> >> {
> >> >> "device": "team0",
> >> >> "runner": {
> >> >> "name": "lacp",
> >> >> "active": 1,
> >> >> "fast_rate": 1
> >> >> },
> >> >> "link_watch": {"name": "ethtool"},
> >> >> "ports": {"eth1": {}, "eth2": {}}
> >> >> }
> >> >>
> >> >> So, is there something wrong with my commands or conf file?
> >> >> Thanks for your help.
> >> >
> >> > You didn't tell what kernel you are using. You need to have
> >> > a kernel with libteam driver enabled.
> >> >
> >> > After booting with that kernel, compile libnl, install it.
> >> > Next compile libteam and install it.
> >> >
> >> > That fixed the same issue here.
> >> > fbl
> >> >
> >
11 years, 6 months
Re: [libteam]teamd_init() failed
by Flavio Leitner
On Fri, 1 Jun 2012 09:43:32 +0800
Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> Hi Flavio,
> What do you mean "I had to clean the 'configure' caches from libnl and
> libteam as well
> in order to get updated binaries."?
> Can you give me some more information ?
yeah, if you modify the environment (adding team driver support in
kernel, for instance) and just run make, the previously cached data
will be used and the tool will report the same error.
You have to run 'make clean' or start with a fresh './autogen.sh'
to update everything.
fbl
>
> Thanks,
> Best Wishes,
> Wenhua
>
>
>
> 2012/6/1 Wenhua Zhang <shiziwen(a)gmail.com>:
> > Hi Flavio,
> > Can you show me your config file and the environment you use?
> > I am confused about how to test the libteam, and I hope you can give
> > me some advice.
> > We plain to use libteam instead of bonding, and before that we need do
> > some tests.
> >
> > Thanks,
> > Best Wishes.
> >
> >
> > 2012/6/1 Flavio Leitner <fbl(a)redhat.com>:
> >> On Thu, 31 May 2012 23:23:42 +0800
> >> Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> >>
> >>> Hi Flavio,
> >>> Thanks for your help.
> >>> I have build the latest net-next kernel and try again.
> >>> There is still some problem, and I will try some other method to fix it.
> >>>
> >>
> >> I did those steps yesterday using net-next and I am now running
> >> active-backup on two different machines.
> >>
> >> I had to clean the 'configure' caches from libnl and libteam as well
> >> in order to get updated binaries.
> >>
> >> fbl
> >>
> >>>
> >>>
> >>> 2012/5/31 Flavio Leitner <fbl(a)redhat.com>:
> >>> > On Thu, 31 May 2012 11:36:43 +0800
> >>> > Wenhua Zhang <shiziwen(a)gmail.com> wrote:
> >>> >
> >>> >> Hi all,
> >>> >> When I try to test the libteam as Example in
> >>> >> https://fedorahosted.org/libteam/wiki/InfrastructureSpecification, I
> >>> >> got errors as blew:
> >>> >> Failed to create team device.
> >>> >> teamd_init() failed.
> >>> >> Failed to start daemon: Operation not supported
> >>> >>
> >>> >> The command is :teamd -f teamd2.conf -D
> >>> >> The conf is :
> >>> >> {
> >>> >> "device": "team0",
> >>> >> "runner": {
> >>> >> "name": "lacp",
> >>> >> "active": 1,
> >>> >> "fast_rate": 1
> >>> >> },
> >>> >> "link_watch": {"name": "ethtool"},
> >>> >> "ports": {"eth1": {}, "eth2": {}}
> >>> >> }
> >>> >>
> >>> >> So, is there something wrong with my commands or conf file?
> >>> >> Thanks for your help.
> >>> >
> >>> > You didn't tell what kernel you are using. You need to have
> >>> > a kernel with libteam driver enabled.
> >>> >
> >>> > After booting with that kernel, compile libnl, install it.
> >>> > Next compile libteam and install it.
> >>> >
> >>> > That fixed the same issue here.
> >>> > fbl
> >>> >
> >>
11 years, 6 months