Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.debian.user > #167201 > unrolled thread
| Started by | Mariusz Gronczewski <xani@efigence.com> |
|---|---|
| First post | 2016-04-29 16:10 +0200 |
| Last post | 2016-05-04 14:00 +0200 |
| Articles | 3 — 3 participants |
Back to article view | Back to linux.debian.user
[Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up Mariusz Gronczewski <xani@efigence.com> - 2016-04-29 16:10 +0200
Re: [Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up Sven Hartge <sven@svenhartge.de> - 2016-04-29 18:00 +0200
Re: [Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up Mariusz Gronczewski <mariusz.gronczewski@efigence.com> - 2016-05-04 14:00 +0200
| From | Mariusz Gronczewski <xani@efigence.com> |
|---|---|
| Date | 2016-04-29 16:10 +0200 |
| Subject | [Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up |
| Message-ID | <rthhg-7lH-25@gated-at.bofh.it> |
I have 2 machines(IBM servers). Both use bnx2 NIC driver but one is VERY
slow to boot thru kernel (2.3s vs 27s)
Config is:
auto bond508
allow-hotplug bond508
iface bond508 inet static
address 1.2.3.4/24
slaves eth0.508 eth1.508
bond-mode active-backup
on "fast" one it "just" works, NIC is initialized first, VLAN
interfaces are created, bond adds them and everything works
on "slow" one bond starts with no interfaces because ifup happens
before NIC driver settled (eth0/1 are not in the system),
it looks like that
Apr 28 17:49:31 blade710 kernel: [ 44.346366] bonding: Ethernet Channel Bonding Driver: v3.7.1 (April 27, 2011)
Apr 28 17:49:31 blade710 kernel: [ 44.421839] bonding: bond508 is being created...
Apr 28 17:49:31 blade710 kernel: [ 44.424930] bonding: bond508: Setting xmit hash policy to layer3+4 (1)
Apr 28 17:49:31 blade710 kernel: [ 44.429511] bonding: bond508: option lacp_rate: mode dependency failed, not supported in mode active-backup(1)
Apr 28 17:49:31 blade710 kernel: [ 44.435247] bonding: bond508: interface eth0.508 does not exist!
Apr 28 17:49:31 blade710 kernel: [ 44.440974] bonding: bond508: interface eth1.508 does not exist!
Apr 28 17:49:31 blade710 kernel: [ 44.444804] IPv6: ADDRCONF(NETDEV_UP): bond508: link is not ready
then NIC boots:
Apr 28 17:49:31 blade710 kernel: [ 44.492469] bnx2 0000:10:00.0: firmware: direct-loading firmware bnx2/bnx2-mips-09-6.2.1b.fw
Apr 28 17:49:31 blade710 kernel: [ 44.494198] bnx2 0000:10:00.0: firmware: direct-loading firmware bnx2/bnx2-rv2p-09-6.0.17.fw
Apr 28 17:49:31 blade710 kernel: [ 44.494243] bnx2 0000:10:00.0: irq 74 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494251] bnx2 0000:10:00.0: irq 75 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494257] bnx2 0000:10:00.0: irq 76 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494263] bnx2 0000:10:00.0: irq 77 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494269] bnx2 0000:10:00.0: irq 78 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494275] bnx2 0000:10:00.0: irq 79 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494281] bnx2 0000:10:00.0: irq 80 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494286] bnx2 0000:10:00.0: irq 81 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.494292] bnx2 0000:10:00.0: irq 82 for MSI/MSI-X
Apr 28 17:49:31 blade710 kernel: [ 44.553877] bnx2 0000:10:00.0 eth0: using MSIX
Apr 28 17:49:31 blade710 kernel: [ 44.553917] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
Apr 28 17:49:31 blade710 kernel: [ 44.610343] bnx2 0000:10:00.0 eth0: NIC SerDes Link is Up, 1000 Mbps full duplex
I *can* make it work on boot if I define eth interfaces separately:
iface eth0.508 inet manual
bond-master bond508
bond-mode active-backup
but that is not exactly pretty solution... is there a way to force
bonding interfaces to get up only after required devices are up ?
--
Mariusz Gronczewski, Administrator
Efigence S. A.
ul. Wołoska 9a, 02-583 Warszawa
T: [+48] 22 380 13 13
F: [+48] 22 380 13 14
E: mariusz.gronczewski@efigence.com
<mailto:mariusz.gronczewski@efigence.com>
[toc] | [next] | [standalone]
| From | Sven Hartge <sven@svenhartge.de> |
|---|---|
| Date | 2016-04-29 18:00 +0200 |
| Subject | Re: [Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up |
| Message-ID | <rtiZI-a2-3@gated-at.bofh.it> |
| In reply to | #167201 |
Mariusz Gronczewski <xani@efigence.com> wrote: > I have 2 machines(IBM servers). Both use bnx2 NIC driver but one is VERY > slow to boot thru kernel (2.3s vs 27s) > Config is: > auto bond508 > allow-hotplug bond508 > iface bond508 inet static > address 1.2.3.4/24 > slaves eth0.508 eth1.508 > bond-mode active-backup Question, separate from your problem: Why do you bond the VLANed interfaces instead of the raw eth0/1 and then VLAN the bond? Is there a special reason you do it that way? I'd expect a configuration like this: auto bond0 iface bond0 inet manual bond_mode active-backup bond_miimon 100 bond_updelay 1000 slaves eth0 eth1 auto bond0.508 iface bond0.508 inet static address 1.2.3.4 netmask 255.255.255.0 Grüße, Sven. -- Sigmentation fault. Core dumped.
[toc] | [prev] | [next] | [standalone]
| From | Mariusz Gronczewski <mariusz.gronczewski@efigence.com> |
|---|---|
| Date | 2016-05-04 14:00 +0200 |
| Subject | Re: [Jessie] Is there a way to delay bonding interface initalization to happen after network driver is up |
| Message-ID | <rv3Dd-38F-25@gated-at.bofh.it> |
| In reply to | #167211 |
On Fri, 29 Apr 2016 17:52:36 +0200, Sven Hartge <sven@svenhartge.de> wrote: > Mariusz Gronczewski <xani@efigence.com> wrote: > > > I have 2 machines(IBM servers). Both use bnx2 NIC driver but one is VERY > > slow to boot thru kernel (2.3s vs 27s) > > > Config is: > > > auto bond508 > > allow-hotplug bond508 > > > iface bond508 inet static > > address 1.2.3.4/24 > > slaves eth0.508 eth1.508 > > bond-mode active-backup > > Question, separate from your problem: Why do you bond the VLANed > interfaces instead of the raw eth0/1 and then VLAN the bond? Is there a > special reason you do it that way? > Bugs in centos/nic drivers, mostly, we're migrating away from it and I want to try to make same config first before playing around with different ways to do it. We have config you've descriped on several c5/6 machines but for some reason it didn't wanted to work in c6 when there was both tagged and untagged vlan on wire. After spending some time trying to figure out why same config in c5 doesn't want to work in c6 I did it that way. We also have few odd servers that need different bond config for different vlans (mostly by having different "active" cards in each to not force all traffic via one switch) -- Mariusz Gronczewski, Administrator Efigence S. A. ul. Wołoska 9a, 02-583 Warszawa T: [+48] 22 380 13 13 F: [+48] 22 380 13 14 E: mariusz.gronczewski@efigence.com <mailto:mariusz.gronczewski@efigence.com>
[toc] | [prev] | [standalone]
Back to top | Article view | linux.debian.user
csiph-web