Tungsten Fabric知识库丨测试2000个vRouter节点部署
作者:Tatsuya Naganawa 译者:TF编译组
由于GCP允许启动多达5k个节点:),因此vRouter的规模测试程序主要针对该平台来介绍。
- 话虽如此,AWS也可以使用相同的程序
第一个目标是2k个vRouter节点,但是据我尝试,这不是最大值,可以通过更多的控制节点或添加CPU/MEM来达到更大的数值。
在GCP中,可以使用多个子网创建VPC,将控制平面节点分配为172.16.1.0/24,而vRouter节点分配为10.0.0.0/9。(默认子网为/12,最多可达到4k个节点)
默认情况下,并非所有实例都可以具有全局IP,因此需要为vRouter节点定义Cloud NAT才能访问Internet。(我将全局IP分配给控制平面节点,因为节点数量不会那么多)
所有节点均由instance-group创建,并且禁用了auto-scaling,并分配了固定编号。所有节点都为了降低成本而配置了抢占模式(vRouter为$0.01/1hr(n1-standard-1),控制平面节点为$0.64/1hr(n1-standard-64))
总体程序描述如下:
- 使用此程序设置control/config x 5,analytics x 3,kube-master x 1。
- https://github.com/tnaganawa/tungstenfabric-docs/blob/master/TungstenFabricPrimer.md#2-tungstenfabric-up-and-running
这一步最多需要30分钟。
使用了2002版本。JVM_EXTRA_OPTS:设置为“-Xms128m -Xmx20g”。
需要补充一点的是,XMPP_KEEPALIVE_SECONDS决定了XMPP的可扩展性,我将其设置为3。因此,在vRouter节点故障之后,control组件需要9秒才能识别出它来。(默认情况下设置为10/30)对于IaaS用例,我认为这是一个中等的选择,但如果这个值需要较低,则需要有更多的CPU。
为了后续使用,这里还创建了虚拟网络vn1(10.1.0.0/12,l2/l3)。
- https://github.com/tnaganawa/tungctl/blob/master/samples.yaml#L3
- 使用此程序设置一个kube-master。
- https://github.com/tnaganawa/tungstenfabric-docs/blob/master/TungstenFabricPrimer.md#kubeadm
这一步最多需要20分钟。
对于cni.yaml,使用以下URL。
- https://github.com/tnaganawa/tungstenfabric-docs/blob/master/multi-kube-master-deployment-cni-tungsten-fabric.yaml
XMPP_KEEPALIVE_SECONDS:“3”已经添加到env中。
由于GCP上的vRouter问题,
- https://github.com/tnaganawa/tungstenfabric-docs/blob/master/TungstenFabricKnowledgeBase.md#vrouters-on-gce-cannot-reach-other-nodes-in-the-same-subnet
vrouter-agent容器已打补丁,并且yaml需要更改。
- name: contrail-vrouter-agentimage: "tnaganawa/contrail-vrouter-agent:2002-latest" ### 这一行改变了
set-label.sh和kubectl apply -f cni.yaml此时已完成。
- 启动vRouter节点,并使用以下命令转储ips。
(for GCP)
gcloud --format="value(networkInterfaces[0].networkIP)" compute instances list(for AWS, this command can be used)
aws ec2 describe-instances --query 'Reservations[*].Instances[*].PrivateIpAddress' --output text | tr '\t' '\n'
这大约需要10-20分钟。
- 在vRouter节点上安装Kubernetes,然后等待它安装vRouter节点。
(/tmp/aaa.pem is the secret key for GCP)
sudo yum -y install epel-release
sudo yum -y install parallel
sudo su - -c "ulimit -n 8192; su - centos"
cat all.txt | parallel -j3500 scp -i /tmp/aaa.pem -o StrictHostKeyChecking=no install-k8s-packages.sh centos@{}:/tmp
cat all.txt | parallel -j3500 ssh -i /tmp/aaa.pem -o StrictHostKeyChecking=no centos@{} chmod 755 /tmp/install-k8s-packages.sh
cat all.txt | parallel -j3500 ssh -i /tmp/aaa.pem -o StrictHostKeyChecking=no centos@{} sudo /tmp/install-k8s-packages.sh###该命令需要最多200个并行执行,如果不执行该命令,会导致kubeadm连接超时
cat all.txt | parallel -j200 ssh -i /tmp/aaa.pem -o StrictHostKeyChecking=no centos@{} sudo kubeadm join 172.16.1.x:6443 --token we70in.mvy0yu0hnxb6kxip --discovery-token-ca-cert-hash sha256:13cf52534ab14ee1f4dc561de746e95bc7684f2a0355cb82eebdbd5b1e9f3634
kubeadm加入大约需要20-30分钟。vRouter安装大约需要40-50分钟。(基于Cloud NAT性能,在VPC中添加docker注册表会缩短docker pull time)。
- 之后,可以使用replica: 2000创建first-containers.yaml,并检查容器之间的ping。要查看BUM行为,vn1也可以与具有2k replica的容器一起使用。
- https://github.com/tnaganawa/tungstenfabric-docs/blob/master/8-leaves-contrail-config.txt#L99
创建容器最多需要15-20分钟。
[config results]2200 instances are created, and 2188 kube workers are available.
(some are restarted and not available, since those instance are preemptive VM)[centos@instance-group-1-srwq ~]$ kubectl get node | wc -l
2188
[centos@instance-group-1-srwq ~]$ When vRouters are installed, some more nodes are rebooted, and 2140 vRouters become available.Every 10.0s: kubectl get pod --all-namespaces | grep contrail | grep Running | wc -l Sun Feb 16 17:25:16 2020
2140After start creating 2k containers, 15 minutes is needed before 2k containers are up.Every 5.0s: kubectl get pod -n myns11 | grep Running | wc -l Sun Feb 16 17:43:06 2020
1927ping between containers works fine:$ kubectl get pod -n myns11
(snip)
vn11-deployment-68565f967b-zxgl4 1/1 Running 0 15m 10.0.6.0 instance-group-3-bqv4
vn11-deployment-68565f967b-zz2f8 1/1 Running 0 15m 10.0.6.16 instance-group-2-ffdq
vn11-deployment-68565f967b-zz8fk 1/1 Running 0 16m 10.0.1.61 instance-group-4-cpb8
vn11-deployment-68565f967b-zzkdk 1/1 Running 0 16m 10.0.2.244 instance-group-3-pkrq
vn11-deployment-68565f967b-zzqb7 0/1 ContainerCreating 0 15m instance-group-4-f5nw
vn11-deployment-68565f967b-zzt52 1/1 Running 0 15m 10.0.5.175 instance-group-3-slkw
vn11-deployment-68565f967b-zztd6 1/1 Running 0 15m 10.0.7.154 instance-group-4-skzk [centos@instance-group-1-srwq ~]$ kubectl exec -it -n myns11 vn11-deployment-68565f967b-zzkdk sh
/ #
/ #
/ # ip -o a
1: lo: mtu 65536 qdisc noqueue qlen 1000\ link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
1: lo inet 127.0.0.1/8 scope host lo\ valid_lft forever preferred_lft forever
1: lo inet6 ::1/128 scope host \ valid_lft forever preferred_lft forever
36: eth0@if37: mtu 1500 qdisc noqueue \ link/ether 02:fd:53:2d:ea:50 brd ff:ff:ff:ff:ff:ff
36: eth0 inet 10.0.2.244/12 scope global eth0\ valid_lft forever preferred_lft forever
36: eth0 inet6 fe80::e416:e7ff:fed3:9cc5/64 scope link \ valid_lft forever preferred_lft forever
/ # ping 10.0.1.61
PING 10.0.1.61 (10.0.1.61): 56 data bytes
64 bytes from 10.0.1.61: seq=0 ttl=64 time=3.635 ms
64 bytes from 10.0.1.61: seq=1 ttl=64 time=0.474 ms
^C
--- 10.0.1.61 ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 0.474/2.054/3.635 ms
/ #There are some XMPP flap .. it might be caused by CPU spike by config-api, or some effect of preemptive VM.
It needs to be investigated further with separating config and control.
(most of other 2k vRouter nodes won't experience XMPP flap though)(venv) [centos@instance-group-1-h26k ~]$ ./contrail-introspect-cli/ist.py ctr nei -t XMPP -c flap_count | grep -v -w 0
+------------+
| flap_count |
+------------+
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
+------------+
(venv) [centos@instance-group-1-h26k ~]$ [BUM tree]Send two multicast packets./ # ping 224.0.0.1
PING 224.0.0.1 (224.0.0.1): 56 data bytes
^C
--- 224.0.0.1 ping statistics ---
2 packets transmitted, 0 packets received, 100% packet loss
/ #
/ # ip -o a
1: lo: mtu 65536 qdisc noqueue qlen 1000\ link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
1: lo inet 127.0.0.1/8 scope host lo\ valid_lft forever preferred_lft forever
1: lo inet6 ::1/128 scope host \ valid_lft forever preferred_lft forever
36: eth0@if37: mtu 1500 qdisc noqueue \ link/ether 02:fd:53:2d:ea:50 brd ff:ff:ff:ff:ff:ff
36: eth0 inet 10.0.2.244/12 scope global eth0\ valid_lft forever preferred_lft forever
36: eth0 inet6 fe80::e416:e7ff:fed3:9cc5/64 scope link \ valid_lft forever preferred_lft forever
/ # That container is on this node.(venv) [centos@instance-group-1-h26k ~]$ ping instance-group-3-pkrq
PING instance-group-3-pkrq.asia-northeast1-b.c.stellar-perigee-161412.internal (10.0.3.211) 56(84) bytes of data.
64 bytes from instance-group-3-pkrq.asia-northeast1-b.c.stellar-perigee-161412.internal (10.0.3.211): icmp_seq=1 ttl=63 time=1.46 msIt sends overlay packet to some other endpoints (not all 2k nodes),[root@instance-group-3-pkrq ~]# tcpdump -nn -i eth0 udp
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
17:48:51.501718 IP 10.0.3.211.57333 > 10.0.0.212.6635: UDP, length 142
17:48:52.501900 IP 10.0.3.211.57333 > 10.0.0.212.6635: UDP, length 142and it eventually reach other containers, going through Edge Replicate tree.[root@instance-group-4-cpb8 ~]# tcpdump -nn -i eth0 udp
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
17:48:51.517306 IP 10.0.1.198.58095 > 10.0.5.244.6635: UDP, length 142
17:48:52.504484 IP 10.0.1.198.58095 > 10.0.5.244.6635: UDP, length 142[resource usage]controller:
CPU usage is moderate and bound by contrail-control process.
If more vRouter nodes need to be added, more controller nodes can be added.- separating config and control also should help to reach further stabilitytop - 17:45:28 up 2:21, 2 users, load average: 7.35, 12.16, 16.33
Tasks: 577 total, 1 running, 576 sleeping, 0 stopped, 0 zombie
%Cpu(s): 14.9 us, 4.2 sy, 0.0 ni, 80.8 id, 0.0 wa, 0.0 hi, 0.1 si, 0.0 st
KiB Mem : 24745379+total, 22992752+free, 13091060 used, 4435200 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 23311113+avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
23930 1999 20 0 9871.3m 5.5g 14896 S 1013 2.3 1029:04 contrail-contro
23998 1999 20 0 5778768 317660 12364 S 289.1 0.1 320:18.42 contrail-dns
13434 polkitd 20 0 33.6g 163288 4968 S 3.3 0.1 32:04.85 beam.smp
26696 1999 20 0 829768 184940 6628 S 2.3 0.1 0:22.14 node 9838 polkitd 20 0 25.4g 2.1g 15276 S 1.3 0.9 45:18.75 java 1012 root 20 0 0 0 0 S 0.3 0.0 0:00.26 kworker/18:1 6293 root 20 0 3388824 50576 12600 S 0.3 0.0 0:34.39 docker-containe 9912 centos 20 0 38.0g 417304 12572 S 0.3 0.2 0:25.30 java
16621 1999 20 0 735328 377212 7252 S 0.3 0.2 23:27.40 contrail-api
22289 root 20 0 0 0 0 S 0.3 0.0 0:00.04 kworker/16:2
24024 root 20 0 259648 41992 5064 S 0.3 0.0 0:28.81 contrail-nodemg
48459 centos 20 0 160328 2708 1536 R 0.3 0.0 0:00.33 top
61029 root 20 0 0 0 0 S 0.3 0.0 0:00.09 kworker/4:2 1 root 20 0 193680 6780 4180 S 0.0 0.0 0:02.86 systemd 2 root 20 0 0 0 0 S 0.0 0.0 0:00.03 kthreadd [centos@instance-group-1-rc34 ~]$ free -htotal used free shared buff/cache available
Mem: 235G 12G 219G 9.8M 3.9G 222G
Swap: 0B 0B 0B
[centos@instance-group-1-rc34 ~]$
[centos@instance-group-1-rc34 ~]$ df -h .
/dev/sda1 10G 5.1G 5.0G 51% /
[centos@instance-group-1-rc34 ~]$ analytics:
CPU usage is moderate and bound by contrail-collector process.
If more vRouter nodes need to be added, more analytics nodes can be added.top - 17:45:59 up 2:21, 1 user, load average: 0.84, 2.57, 4.24
Tasks: 515 total, 1 running, 514 sleeping, 0 stopped, 0 zombie
%Cpu(s): 3.3 us, 1.3 sy, 0.0 ni, 95.4 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem : 24745379+total, 24193969+free, 3741324 used, 1772760 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 24246134+avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 6334 1999 20 0 7604200 958288 10860 S 327.8 0.4 493:31.11 contrail-collec 4904 polkitd 20 0 297424 271444 1676 S 14.6 0.1 10:42.34 redis-server 4110 root 20 0 3462120 95156 34660 S 1.0 0.0 1:21.32 dockerd 9 root 20 0 0 0 0 S 0.3 0.0 0:04.81 rcu_sched 29 root 20 0 0 0 0 S 0.3 0.0 0:00.05 ksoftirqd/4 8553 centos 20 0 160308 2608 1536 R 0.3 0.0 0:00.07 top 1 root 20 0 193564 6656 4180 S 0.0 0.0 0:02.77 systemd 2 root 20 0 0 0 0 S 0.0 0.0 0:00.03 kthreadd 4 root 0 -20 0 0 0 S 0.0 0.0 0:00.00 kworker/0:0H 5 root 20 0 0 0 0 S 0.0 0.0 0:00.77 kworker/u128:0 6 root 20 0 0 0 0 S 0.0 0.0 0:00.17 ksoftirqd/0 7 root rt 0 0 0 0 S 0.0 0.0 0:00.38 migration/0 [centos@instance-group-1-n4c7 ~]$ free -htotal used free shared buff/cache available
Mem: 235G 3.6G 230G 8.9M 1.7G 231G
Swap: 0B 0B 0B
[centos@instance-group-1-n4c7 ~]$
[centos@instance-group-1-n4c7 ~]$ df -h .
/dev/sda1 10G 3.1G 6.9G 32% /
[centos@instance-group-1-n4c7 ~]$ kube-master:
CPU usage is small.top - 17:46:18 up 2:22, 2 users, load average: 0.92, 1.32, 2.08
Tasks: 556 total, 1 running, 555 sleeping, 0 stopped, 0 zombie
%Cpu(s): 1.2 us, 0.5 sy, 0.0 ni, 98.2 id, 0.1 wa, 0.0 hi, 0.1 si, 0.0 st
KiB Mem : 24745379+total, 23662128+free, 7557744 used, 3274752 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 23852964+avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 5177 root 20 0 3605852 3.1g 41800 S 78.1 1.3 198:42.92 kube-apiserver 5222 root 20 0 10.3g 633316 410812 S 55.0 0.3 109:52.52 etcd 5198 root 20 0 846948 651668 31284 S 8.3 0.3 169:03.71 kube-controller 5549 root 20 0 4753664 96528 34260 S 5.0 0.0 4:45.54 kubelet 3493 root 20 0 4759508 71864 16040 S 0.7 0.0 1:52.67 dockerd-current 5197 root 20 0 500800 307056 17724 S 0.7 0.1 6:43.91 kube-scheduler
19933 centos 20 0 160340 2648 1528 R 0.7 0.0 0:00.07 top 1083 root 0 -20 0 0 0 S 0.3 0.0 0:20.19 kworker/0:1H
35229 root 20 0 0 0 0 S 0.3 0.0 0:15.08 kworker/0:2 1 root 20 0 193808 6884 4212 S 0.0 0.0 0:03.59 systemd 2 root 20 0 0 0 0 S 0.0 0.0 0:00.03 kthreadd 4 root 0 -20 0 0 0 S 0.0 0.0 0:00.00 kworker/0:0H 5 root 20 0 0 0 0 S 0.0 0.0 0:00.55 kworker/u128:0 6 root 20 0 0 0 0 S 0.0 0.0 0:01.51 ksoftirqd/0 [centos@instance-group-1-srwq ~]$ free -htotal used free shared buff/cache available
Mem: 235G 15G 217G 121M 3.2G 219G
Swap: 0B 0B 0B
[centos@instance-group-1-srwq ~]$ [centos@instance-group-1-srwq ~]$ df -h /
/dev/sda1 10G 4.6G 5.5G 46% /
[centos@instance-group-1-srwq ~]$ [centos@instance-group-1-srwq ~]$ free -htotal used free shared buff/cache available
Mem: 235G 15G 217G 121M 3.2G 219G
Swap: 0B 0B 0B
[centos@instance-group-1-srwq ~]$ [centos@instance-group-1-srwq ~]$ df -h /
/dev/sda1 10G 4.6G 5.5G 46% /
[centos@instance-group-1-srwq ~]$
附:erm-vpn
启用erm-vpn后,vrouter会将多播流量发送到最多4个节点上,以避免入口(ingress)复制到所有节点。控制器以生成树的方式,将多播数据包发送到所有节点。
- https://tools.ietf.org/html/draft-marques-l3vpn-mcast-edge-00
- https://review.opencontrail.org/c/Juniper/contrail-controller/+/256
为了说明此功能,我创建了一个具有20个kubernete worker的集群,并部署了20个副本。
- 没有使用default-k8s-pod-network,因为它是仅支持l3转发的模式。这里手动定义了vn1(10.0.1.0/24)
在此设置中,以下的命令将转储下一跳(下一跳将发送overlay BUM流量)。
- vrf 2在每个worker节点上都有vn1的路由
- all.txt上具有20个节点的IP
- 当BUM数据包从“ Introspect Host”上的容器发送时,它们以单播overlay的方式发送到“dip”地址
[root@ip-172-31-12-135 ~]# for i in $(cat all.txt); do ./contrail-introspect-cli/ist.py --host $i vr route -v 2 --family layer2 ff:ff:ff:ff:ff:ff -r | grep -w -e dip -e Introspect | sort -r | uniq ; done
Introspect Host: 172.31.15.27dip: 172.31.7.18
Introspect Host: 172.31.4.249dip: 172.31.9.151dip: 172.31.9.108dip: 172.31.8.233dip: 172.31.2.127dip: 172.31.10.233
Introspect Host: 172.31.14.220dip: 172.31.7.6
Introspect Host: 172.31.8.219dip: 172.31.3.56
Introspect Host: 172.31.7.223dip: 172.31.3.56
Introspect Host: 172.31.2.127dip: 172.31.7.6dip: 172.31.7.18dip: 172.31.4.249dip: 172.31.3.56
Introspect Host: 172.31.14.255dip: 172.31.7.6
Introspect Host: 172.31.7.6dip: 172.31.2.127dip: 172.31.14.255dip: 172.31.14.220dip: 172.31.13.115dip: 172.31.11.208
Introspect Host: 172.31.10.233dip: 172.31.4.249
Introspect Host: 172.31.15.232dip: 172.31.7.18
Introspect Host: 172.31.9.108dip: 172.31.4.249
Introspect Host: 172.31.8.233dip: 172.31.4.249
Introspect Host: 172.31.8.206dip: 172.31.3.56
Introspect Host: 172.31.7.142dip: 172.31.3.56
Introspect Host: 172.31.15.210dip: 172.31.7.18
Introspect Host: 172.31.11.208dip: 172.31.7.6
Introspect Host: 172.31.13.115dip: 172.31.9.151
Introspect Host: 172.31.7.18dip: 172.31.2.127dip: 172.31.15.27dip: 172.31.15.232dip: 172.31.15.210
Introspect Host: 172.31.3.56dip: 172.31.8.219dip: 172.31.8.206dip: 172.31.7.223dip: 172.31.7.142dip: 172.31.2.127
Introspect Host: 172.31.9.151dip: 172.31.13.115
[root@ip-172-31-12-135 ~]#
举例来说,我尝试从worker 172.31.7.18上的一个容器向一个多播地址($ ping 224.0.0.1)发送ping信息,它向dip列表中的计算节点发送了4个数据包。
[root@ip-172-31-7-18 ~]# tcpdump -nn -i eth0 -v udp port 663515:02:29.883608 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], proto UDP (17), length 170)172.31.7.18.63685 > 172.31.2.127.6635: UDP, length 142
15:02:29.883623 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], proto UDP (17), length 170)172.31.7.18.63685 > 172.31.15.27.6635: UDP, length 142
15:02:29.883626 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], proto UDP (17), length 170)172.31.7.18.63685 > 172.31.15.210.6635: UDP, length 142
15:02:29.883629 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], proto UDP (17), length 170)172.31.7.18.63685 > 172.31.15.232.6635: UDP, length 142
未定义为直接下一跳的其它节点(172.31.7.223)也接收了多播数据包,尽管它的延迟有所增加。
- 在这种情况下,需要额外增加2个跃点:172.31.7.18-> 172.31.2.127-> 172.31.3.56->
172.31.7.223 [root@ip-172-31-7-223 ~]# tcpdump -nn -i eth0 -v udp port 6635
15:02:29.884070 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], proto UDP (17), length 170)172.31.3.56.56541 > 172.31.7.223.6635: UDP, length 142
原文链接:
https://github.com/tnaganawa/tungstenfabric-docs/blob/master/TungstenFabricKnowledgeBase.md
往期精选
Tungsten Fabric知识库丨vRouter内部运行探秘
Tungsten Fabric知识库丨更多组件内部探秘
Tungsten Fabric知识库丨 构建、安装与公有云部署
Tungsten Fabric入门宝典系列文章——
1.首次启动和运行指南
2. TF组件的七种“武器”
3. 编排器集成
4.关于安装的那些事(上)
5.关于安装的那些事(下)
6.主流监控系统工具的集成
7.开始第二天的工作
8.8个典型故障及排查Tips
9.关于集群更新的那些事
10.说说L3VPN及EVPN集成
11.关于服务链、BGPaaS及其它
12.关于多集群和多数据中心
13.多编排器用法及配置
Tungsten Fabric知识库丨测试2000个vRouter节点部署相关推荐
- Tungsten Fabric知识库丨关于OpenStack、K8s、CentOS安装问题的补充
作者:Tatsuya Naganawa 译者:TF编译组 多kube-master部署 3个Tungsten Fabric控制器节点:m3.xlarge(4 vcpu)-> c3.4xlarge ...
- Tungsten Fabric知识库丨vRouter内部运行探秘
原文链接: https://github.com/tnaganawa/tungstenfabric-docs/blob/master/TungstenFabricKnowledgeBase.md 作者 ...
- TF Live 直播回放丨Frank Wu:当OpenStack遇到Tungsten Fabric
10岁的OpenStack,已经是开源IaaS世界里的"成年人",自从遇到开源SDN小伙伴Tungsten Fabric,两人便成为闯荡混合多云世界的好搭档. 5月26日,在TF中 ...
- 在K8s上轻松部署Tungsten Fabric的两种方式
第一种:在AWS的K8s上部署TF 首先介绍下如何在AWS上使用Kubernetes编排的Tungsten Fabric集群部署沙盒,15分钟就可以搞定.Tungsten Fabric集群由部署节点. ...
- TF实战丨使用Vagrant安装Tungsten Fabric
本文为苏宁网络架构师陈刚的原创文章. 01 准备测试机 在16G的笔记本没跑起来,就干脆拼凑了一台游戏工作室级别的机器:双路E5-2860v3 CPU,24核48线程,128G DDR4 ECC内存, ...
- Tungsten Fabric SDN — Netronome Agilio SmartNIC vRouter
目录 文章目录 目录 Netronome Agilio vRouter 软件架构 SmartNIC SR-IOV / XVIO 部署架构 性能测试 OpenStack 集成 参考文档 Netronom ...
- Tungsten Fabric入门宝典丨8个典型故障及排查Tips
Tungsten Fabric入门宝典系列文章,来自技术大牛倾囊相授的实践经验,由TF中文社区为您编译呈现,旨在帮助新手深入理解TF的运行.安装.集成.调试等全流程.如果您有相关经验或疑问,欢迎与我们 ...
- Tungsten Fabric SDN — 与 Kubernetes 的集成部署(CN)
目录 文章目录 目录 部署架构 软件版本 部署 Kubernetes & TF 基础环境设置 创建 Deployment instance 执行 Playbooks 环境检查 配置记录 Con ...
- Tungsten Fabric如何实现路由的快速收敛?收敛速度有多快?
在发布R2008版本之前,Tungsten Fabric无法同时提供南北向和东西向流量的快速收敛. 实际上,对于南北流量来说,是可以实现快速收敛的,但机制在Tungsten Fabric的逻辑之外: ...
最新文章
- shell逻辑判断式与表达式
- 查看python安装路径
- 【NOIP2007】【Luogu1094】纪念品分组(贪心,乘船问题)
- 【python初识】数据和对象
- leetcode 752. 打开转盘锁 c代码
- 如何通过标签体系,打造精细化运营?
- Android快速开发系列 10个常用工具类
- mysql5.7.14多实例安装
- SAP产品和3D渲染技术的结合-使用JavaScript的开源3D渲染库实现
- windsns社交门户系统源码v1.0-掌上移动社交类SNS网站系统源码
- 视频教程-ArcGIS开发arcpy教程-其他
- DELPHI 字符转16进制、16进制转字符
- echarts 为x轴、y轴添加滚动条
- 书架html特效代码,原生JS写的一个书架式的图片缩放滚动展示特效代码
- Linux云计算虚拟化-使用rancher搭建k8s集群并发布电商网站
- [HTML] kbd 标签
- 基于STM32蓝牙控制的app智能台灯设计
- Wish卖家福利:PayPal通过WorldFirst提现0费率!
- fisco bcos使用在线IDE chainIDE教程,只要有网络就可以写智能合约!
- OriginPro,如何把软件Origin切换变成中文显示
热门文章
- Codeforces Round #416 (Div. 2)
- 共享网站服务器的优缺点,网站服务器的带宽如何选择?独享带宽还是共享带宽?...
- Windows下安装配置PHP
- 《顶级摄影器材》系列丛书首发式上海隆重举办
- 计算机编程语言(1)
- Deep Reinforcement Learning入门 - DQN/Policy Gradient实现LunarLander-v2
- 洛谷P3480 KAM-Pebbles
- Nmap手册(转自http://www.nmap.com.cn/doc/manual.shtm)
- 小邵教你玩转Typescript、ts版React全家桶脚手架
- 台北故宫博物院是台湾游的“金字招牌”