问题导读:
1.怎样使用arp_responder ?
3. ARP Responder
3.1 升级 OVS
1
2
3
|
uname
-r
.
/configure
--with-linux=
/lib/modules/3
.13.0-51-generic
/build
make
&&
make
install
|
1
2
3
4
|
root@compute2:
/home/s1
# ovs-vsctl --version
ovs-vsctl (Open vSwitch) 2.3.2
Compiled Jul 12 2015 09:09:42
DB Schema 7.6.2
|
1
2
|
rm
/etc/openvswitch/conf
.db (老的db要删除掉,否则会报错)
ovsdb-tool create
/etc/openvswitch/conf
.db vswitchd
/vswitch
.ovsschema
|
1
2
3
|
cp
/usr/local/bin/ovs-
*
/usr/bin
ovsdb-server
/etc/openvswitch/conf
.db -vconsole:emer -vsyslog:err -vfile:info --remote=punix:
/usr/local/var/run/openvswitch/db
.sock --private-key=db:Open_vSwitch,SSL,private_key --certificate=db:Open_vSwitch,SSL,certificate --bootstrap-ca-cert=db:Open_vSwitch,SSL,ca_cert --no-chdir --log-
file
=
/var/log/openvswitch/ovsdb-server
.log --pidfile=
/var/run/openvswitch/ovsdb-server
.pid --detach --monitor
ovs-vswitchd unix:
/usr/local/var/run/openvswitch/db
.sock -vconsole:emer -vsyslog:err -vfile:info --mlockall --no-chdir --log-
file
=
/var/log/openvswitch/ovs-vswitchd
.log --pidfile=
/var/run/openvswitch/ovs-vswitchd
.pid --detach --monitor
|
3.2 ARP Responder
# ARP broadcast-ed request go to the local ARP_RESPONDER table to be locally resolved
table=2, n_packets=0, n_bytes=0, idle_age=3, priority=1,arp,dl_dst=ff:ff:ff:ff:ff:ff actions=resubmit(,21)
# If none of the ARP entries correspond to the requested IP, the broadcast-ed packet is resubmitted to the flooding table
table=21, n_packets=0, n_bytes=0, idle_age=4, priority=0 actions=resubmit(,22)
如果下面第 (3)步增加的 flow rule 都处理不了这条 request,那么转到table 22 去 flood 到所有端口。
br.add_flow(table=21, priority=1, proto='arp', dl_vlan=local_vid, nw_dst= ip, actions=actions)
1
2
3
4
5
6
7
8
|
ARP_RESPONDER_ACTIONS
=
(
'move:NXM_OF_ETH_SRC[]->NXM_OF_ETH_DST[],'
'mod_dl_src:%(mac)s,'
'load:0x2->NXM_OF_ARP_OP[],'
'move:NXM_NX_ARP_SHA[]->NXM_NX_ARP_THA[],'
'move:NXM_OF_ARP_SPA[]->NXM_OF_ARP_TPA[],'
'load:%(mac)#x->NXM_NX_ARP_SHA[],'
'load:%(ip)#x->NXM_OF_ARP_SPA[],'
'in_port'
)
|
(4)table 21 的处理过程
4. L2 population
4.1 原理
{'status': 'DOWN/BUILD/ACTIVE', 'binding:host_id': u'compute1', 'allowed_address_pairs': [], 'extra_dhcp_opts': [], 'device_owner': u'compute:nova', 'binding:profile': {}, 'fixed_ips': [{'subnet_id': u'4ec65731-35a5-4637-a59b-a9f2932099f1', 'ip_address': u'81.1.180.15'}], 'id': u'1167e9ac-e10f-4cf5-bd09-6649eab38b32', 'security_groups': [u'f5377a66-803d-481b-b4c3-a6631e8ab456'], 'device_id': u'30580ea7-c456-416b-a01e-0fe645edf5dc', 'name': u'', 'admin_state_up': True, 'network_id': u'86c0d29b-4880-4739-bd68-eb3c392f5099', 'tenant_id': u'74c8ada23a3449f888d9e19b76d13aab', 'binding:vif_details': {u'port_filter': True, u'ovs_hybrid_plug': True}, 'binding:vnic_type': u'normal', 'binding:vif_type': u'ovs', 'mac_address': u'fa:16:3e:4f:59:9d'}
- update_port_postcommit (down to active) -> _update_port_up -> add_fdb_entries -> fdb_add -> fdb_add_tun -> setup_tunnel_port (如果 tunnel port 不存在,则创建 tunnel port), add_fdb_flow -> add FLOOD_TO_TUN flow (如果是 Flood port,则将端口添加到 Flood output ports); setup_entry_for_arp_reply('add'。如果不是 Flood port,那么 添加 ARP Responder entry (MAC -> IP)) 以及 add UCAST_TO_TUN flow Unicast Flow entry (MAC -> Tunnel port number)。
- update_port_postcommit (active to down) -> _update_port_down -> remove_fdb_entries
- delete_port_postcommit (active to down) -> _update_port_down -> remove_fdb_entries -> fdb_remove -> fdb_remove_tun -> cleanup_tunnel_port, del_fdb_flow -> mod/del FLOOD_TO_TUN flow; setup_entry_for_arp_reply ('remove'), delete UCAST_TO_TUN flow
- update_port_postcommit (fixed ip changed) -> _fixed_ips_changed -> update_fdb_entries
- Tunnel port
- FLOOD_TO_TUN (table 22)flow
- ARP responder flow
- UCAST_TO_TUN (table 20) flow
4.2 过程实验
- 通知 compute1: {'segment_id': 6L, 'ports': {u'10.0.1.21': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:87:40:f3', u'81.1.180.1']]}, 'network_type': u'gre'}}
- 通知所有 agent: {'segment_id': 6L, 'ports': {u'10.0.1.31': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:b3:e7:7a', u'81.1.180.8']]}, 'network_type': u'gre'}}
- 添加和网络节点的tunnel options: {df_default="true", in_key=flow, local_ip="10.0.1.31", out_key=flow, remote_ip="10.0.1.21"}
- 添加到网段网关的 Unicast flow:table=20, n_packets=0, n_bytes=0, idle_age=130, priority=2,dl_vlan=2,dl_dst=fa:16:3e:87:40:f3 actions=strip_vlan,set_tunnel:0x6,output:4
- 增加网段 81.1.180.1 网关的 ARP flows:table=21, n_packets=0, n_bytes=0, idle_age=130, priority=1,arp,dl_vlan=2,arp_tpa=81.1.180.1 actions=move:NXM_OF_ETH_SRC[]->NXM_OF_ETH_DST[],mod_dl_src:fa:16:3e:87:40:f3,load:0x2->NXM_OF_ARP_OP[],move:NXM_NX_ARP_SHA[]->NXM_NX_ARP_THA[],move:NXM_OF_ARP_SPA[]->NXM_OF_ARP_TPA[],load:0xfa163e8740f3->NXM_NX_ARP_SHA[],load:0x5101b401->NXM_OF_ARP_SPA[],IN_PORT
- 修改 Flood flow
- 通知 compute 2 : {'segment_id': 6L, 'ports': {u'10.0.1.31': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:b3:e7:7a', u'81.1.180.8']], u'10.0.1.21': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:87:40:f3', u'81.1.180.1']]}, 'network_type': u'gre'}}
- 通知所有 agent: {'segment_id': 6L, 'ports': {u'10.0.1.39': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:73:49:41', u'81.1.180.9']]}, 'network_type': u'gre'}
- 建立 tunnel(ID 5): {df_default="true", in_key=flow, local_ip="10.0.1.31", out_key=flow, remote_ip="10.0.1.39"}
- 增加 arp responder flow(compute2 上新的虚机 IP -> MAC):table=21, n_packets=0, n_bytes=0, idle_age=79, priority=1,arp,dl_vlan=2,arp_tpa=81.1.180.9 actions=move:NXM_OF_ETH_SRC[]->NXM_OF_ETH_DST[],mod_dl_src:fa:16:3e:73:49:41,load:0x2->NXM_OF_ARP_OP[],move:NXM_NX_ARP_SHA[]->NXM_NX_ARP_THA[],move:NXM_OF_ARP_SPA[]->NXM_OF_ARP_TPA[],load:0xfa163e734941->NXM_NX_ARP_SHA[],load:0x5101b409->NXM_OF_ARP_SPA[],IN_PORT
- 增加 unicast flow (新虚机的 MAC -> Tunnel port):table=20, n_packets=0, n_bytes=0, idle_age=79, priority=2,dl_vlan=2,dl_dst=fa:16:3e:73:49:41 actions=strip_vlan,set_tunnel:0x6,output:5
- 添加新的 Tunnel port 到 Flood flow:table=22, n_packets=13, n_bytes=1717, idle_age=255, hard_age=78, dl_vlan=2 actions=strip_vlan,set_tunnel:0x6,output:5,output:4
- 建立和计算节点以及compute1的tunnel:options: {df_default="true", in_key=flow, local_ip="10.0.1.39", out_key=flow, remote_ip="10.0.1.21"},options: {df_default="true", in_key=flow, local_ip="10.0.1.39", out_key=flow, remote_ip="10.0.1.31"}
- 增加 ARP flow(compute 1 上的虚机的 MAC -> IP):table=21, n_packets=0, n_bytes=0, idle_age=268, priority=1,arp,dl_vlan=2,arp_tpa=81.1.180.8 actions=move:NXM_OF_ETH_SRC[]->NXM_OF_ETH_DST[],mod_dl_src:fa:16:3e:b3:e7:7a,load:0x2->NXM_OF_ARP_OP[],move:NXM_NX_ARP_SHA[]->NXM_NX_ARP_THA[],move:NXM_OF_ARP_SPA[]->NXM_OF_ARP_TPA[],load:0xfa163eb3e77a->NXM_NX_ARP_SHA[],load:0x5101b408->NXM_OF_ARP_SPA[],IN_PORT
- 增加 Unicast flow (compute 1 上的虚机 MAC -> Tunnel port):table=20, n_packets=0, n_bytes=0, idle_age=268, priority=2,dl_vlan=2,dl_dst=fa:16:3e:b3:e7:7a actions=strip_vlan,set_tunnel:0x6,output:4
-
增加 ARP flow(新虚机的网关的 MAC -> IP) table=21, n_packets=0, n_bytes=0, idle_age=268, priority=1,arp,dl_vlan=2,arp_tpa=81.1.180.1 actions=move:NXM_OF_ETH_SRC[]->NXM_OF_ETH_DST[],mod_dl_src:fa:16:3e:87:40:f3,load:0x2->NXM_OF_ARP_OP[],move:NXM_NX_ARP_SHA[]->NXM_NX_ARP_THA[],move:NXM_OF_ARP_SPA[]->NXM_OF_ARP_TPA[],load:0xfa163e8740f3->NXM_NX_ARP_SHA[],load:0x5101b401->NXM_OF_ARP_SPA[],IN_PORT
- 修改 Flood flow(添加到 Compute 1 的 port):table=22, n_packets=13, n_bytes=1717, idle_age=128, dl_vlan=2 actions=strip_vlan,set_tunnel:0x6,output:5,output:4
- 通知所有 agent: {'segment_id': 6L, 'ports': {u'10.0.1.31': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:b3:e7:7a', u'81.1.180.8']]}, 'network_type': u'gre'}
- 因为没有别的虚机了,删除所有 tunnel ports
- 修改或者删除 ARP, Unicast 和 Flood flows
- 删除了 compute1 的 tunnel
- 删除该虚机对应的 ARP flow
- 通知 compute 1: {u'e2022937-ec2a-467a-8cf1-f642a3f777b6': {'segment_id': 4L, 'ports': {u'10.0.1.21': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:90:e5:50', u'91.1.180.1'], [u'fa:16:3e:17:c9:26', u'90.1.180.1'], [u'fa:16:3e:69:92:30', u'90.1.180.3'], [u'fa:16:3e:69:92:30', u'91.1.180.2']]}, 'network_type': u'gre'}}
- 通知所有 agent:{u'e2022937-ec2a-467a-8cf1-f642a3f777b6': {'segment_id': 4L, 'ports': {u'10.0.1.31': [['00:00:00:00:00:00', '0.0.0.0'], [u'fa:16:3e:e9:ee:0c', u'91.1.180.9']]}, 'network_type': u'gre'}}
- 虚机在收到 fannout FDB entries 后,检查其中每个 port 的 network_id(即 “segment_id”)。如果本机上有该 network 内的 port,那么就处理 entries 中的 “ports”部分;否则,不处理该 entries。
- 因此,当计算节点上没有运行任何虚机时,不会建立任何 tunnel。如果两个虚机上有相同网络内的虚机,那么建立会建立 tunnel。
- 这种机制能实时建立 tunnel port,Flood entry (创建 Tunnel port 同时添加到 Flood output ports 列表), Unicast flow (虚机和网关 MAC -> Tunnel port) 和 ARP Responder entry (虚机和网关 MAC -> IP)。下图中的蓝色部分的流表都会被及时更新。
- Neutron server 在端口创建/删除/修改时,如果是该节点上的第一个虚机,首先发送直接消息;然后发通知消息给所有的计算和网络节点。
4.3 性能
4.3.1 MQ 性能问题
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
|
def
__init__(
self
, topic
=
topics.AGENT):
super
(L2populationAgentNotifyAPI,
self
).__init__(
topic
=
topic, default_version
=
self
.BASE_RPC_API_VERSION)
self
.topic_l2pop_update
=
topics.get_topic_name(topic, topics.L2POPULATION, topics.UPDATE)
def
_notification_fanout(
self
, context, method, fdb_entries):
self
.fanout_cast(context,
self
.make_msg(method, fdb_entries
=
fdb_entries), topic
=
self
.topic_l2pop_update)
def
_notification_host(
self
, context, method, fdb_entries, host):
self
.cast(context,
self
.make_msg(method, fdb_entries
=
fdb_entries), topic
=
'%s.%s'
%
(
self
.topic_l2pop_update, host))
def
add_fdb_entries(
self
, context, fdb_entries, host
=
None
):
if
fdb_entries:
if
host:
self
._notification_host(context,
'add_fdb_entries'
,fdb_entries, host)
#cast 给指定 host
else
:
self
._notification_fanout(context,
'add_fdb_entries'
, fdb_entries)
#fanout 给所有计算和网络节点
|
4.3.2 大规模网络环境中节点上的 OpenFlow flows 过多