Purpose: Docker network tshooting can be difficult for network engineers. With proper understanding of how Docker networking works and the right set of tools, you can troubleshoot and resolve these networking issues. The netshoot container has a set of powerful networking tshooting tools that can be used to troubleshoot Docker networking issues.
Network Namespaces: Before starting to use this tool, it's important to go over one key topic: Network Namespaces. Network namespaces provide isolation of the system resources associated with networking. Docker uses network and other type of namespaces (pid,mount,user..etc) to create an isolated environment for each container. Everything from interfaces, routes, and IPs is completely isolated within the network namespace of the container.
Cool thing about namespaces is that you can switch between them. You can enter a different container's network namespace, perform some troubleshooting on its network's stack with tools that aren't even installed on that container. Additionally, netshoot can be used to troubleshoot the host itself by using the host's network namespace. This allows you to perform any troubleshooting without installing any new packages directly on the host or your application's package.
- Container's Network Namespace: If you're having networking issues with your application's container, you can launch
netshootwith that container's network namespace like this :
$ docker run -it --net container:<container_name> nicolaka/netshoot
- Host's Network Namespace: If you think the networking issue is on the host itself, you can launch
netshootwith that host's network namespace. This is how:
$ docker run -it --net host nicolaka/netshoot
- Network's Network Namespace: If you want to troubleshoot a Docker network, you can enter the network's namespace using
nsenter. This is explained in thensentersection below.
Network Problems: Many network issues could result in application performance degradations. Some of those issues could be related to the underlying networking infrastructure(underlay). Others could be related to misconfiguration at the host or Docker level. Let's take a look at common networking issues:
- latency
- routing
- DNS resolution
- firewall
- incomplete ARPs
To troubleshoot these issues, netshoot includes a set of powerful tools as recommended by this diagram.
Included Packages: The following packages are included in netshoot. We'll go over some with some sample use-cases.
- iperf
- tcpdump
- netstat
- iftop
- drill
- netcat-openbsd
- iproute2
- util-linux(nsenter)
- bridge-utils
- iputils
- curl
- ethtool
- nmap
- ipvs
- ngrep
Purpose : test networking performance between two containers/hosts.
Create Overlay network:
$docker network create -d overlay perf-test
Launch two containers:
🐳 → docker service create --name perf-test-a --network perf-test nicolaka/netshoot iperf -s -p 9999
7dkcckjs0g7b4eddv8e5ez9nv
🐳 → docker service create --name perf-test-b --network perf-test nicolaka/netshoot iperf -c perf-test-a -p 9999
2yb6fxls5ezfnav2z93lua8xl
🐳 → docker service ls
ID NAME REPLICAS IMAGE COMMAND
2yb6fxls5ezf perf-test-b 1/1 nicolaka/netshoot iperf -c perf-test-a -p 9999
7dkcckjs0g7b perf-test-a 1/1 nicolaka/netshoot iperf -s -p 9999
🐳 → docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
ce4ff40a5456 nicolaka/netshoot:latest "iperf -s -p 9999" 31 seconds ago Up 30 seconds perf-test-a.1.bil2mo8inj3r9nyrss1g15qav
🐳 → docker logs ce4ff40a5456
------------------------------------------------------------
Server listening on TCP port 9999
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[ 4] local 10.0.3.3 port 9999 connected with 10.0.3.5 port 35102
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-10.0 sec 32.7 GBytes 28.1 Gbits/sec
[ 5] local 10.0.3.3 port 9999 connected with 10.0.3.5 port 35112
tcpdump is a powerful and common packet analyzer that runs under the command line. It allows the user to display TCP/IP and other packets being transmitted or received over an attached network interface.
# Continuing on the iperf example. Let's lauch netshoot with perf-test-a's container network namespace.
🐳 → docker run -it --net container:perf-test-a.1.0qlf1kaka0cq38gojf7wcatoa nicolaka/netshoot
# Capturing packets on eth0 and tcp port 9999.
/ # tcpdump -i eth0 port 9999 -c 1 -Xvv
tcpdump: listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
23:14:09.771825 IP (tos 0x0, ttl 64, id 60898, offset 0, flags [DF], proto TCP (6), length 64360)
10.0.3.5.60032 > 0e2ccbf3d608.9999: Flags [.], cksum 0x1563 (incorrect -> 0x895d), seq 222376702:222441010, ack 3545090958, win 221, options [nop,nop,TS val 2488870 ecr 2488869], length 64308
0x0000: 4500 fb68 ede2 4000 4006 37a5 0a00 0305 E..h..@.@.7.....
0x0010: 0a00 0303 ea80 270f 0d41 32fe d34d cb8e ......'..A2..M..
0x0020: 8010 00dd 1563 0000 0101 080a 0025 fa26 .....c.......%.&
0x0030: 0025 fa25 0000 0000 0000 0001 0000 270f .%.%..........'.
0x0040: 0000 0000 0000 0000 ffff d8f0 3435 3637 ............4567
0x0050: 3839 3031 3233 3435 3637 3839 3031 3233 8901234567890123
0x0060: 3435 3637 3839 3031 3233 3435 3637 3839 4567890123456789
0x0070: 3031 3233 3435 3637 3839 3031 3233 3435 0123456789012345
0x0080: 3637 3839 3031 3233 3435 3637 3839 3031 6789012345678901
0x0090: 3233 3435 3637 3839 3031 3233 3435 3637 2345678901234567
0x00a0: 3839 3031 3233 3435 3637 3839 3031 3233 8901234567890123
0x00b0: 3435 3637 3839 3031 3233 3435 3637 3839 4567890123456789
0x00c0: 3031 3233 3435 3637 3839 3031 3233 3435 0123456789012345
0x00d0: 3637 3839 3031 3233 3435 3637 3839 3031 6789012345678901
0x00e0: 3233 3435 3637 3839 3031 3233 3435 3637 2345678901234567
0x00f0: 3839 3031 3233 3435 3637 3839 3031 3233 8901234567890123
0x0100: 3435 3637 3839 3031 3233 3435 3637 3839 4567890123456789
More info on tcpdump can be found here.
Purpose: netstat is a useful tool for checking your network configuration and activity.
Continuing on from iperf example. Let's use netstat to confirm that it's listening on port 9999.
🐳 → docker run -it --net container:perf-test-a.1.0qlf1kaka0cq38gojf7wcatoa nicolaka/netshoot
/ # netstat -tulpn
Active Internet connections (only servers)
Proto Recv-Q Send-Q Local Address Foreign Address State PID/Program name
tcp 0 0 127.0.0.11:46727 0.0.0.0:* LISTEN -
tcp 0 0 0.0.0.0:9999 0.0.0.0:* LISTEN -
udp 0 0 127.0.0.11:39552 0.0.0.0:* -
Purpose: iftop does for network usage what top does for CPU usage. It listens to network traffic on a named interface and displays a table of current bandwidth usage by pairs of hosts.
Continuing the iperf example.
→ docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
ce4ff40a5456 nicolaka/netshoot:latest "iperf -s -p 9999" 5 minutes ago Up 5 minutes perf-test-a.1.bil2mo8inj3r9nyrss1g15qav
🐳 → docker run -it --net container:perf-test-a.1.bil2mo8inj3r9nyrss1g15qav nicolaka/netshoot iftop -i eth0
Purpose: drill is a tool to designed to get all sorts of information out of the DNS.
Continuing the iperf example, we'll use drill to understand how services' DNS is resolved in Docker.
🐳 → docker run -it --net container:perf-test-a.1.bil2mo8inj3r9nyrss1g15qav nicolaka/netshoot drill -V 5 perf-test-b
;; ->>HEADER<<- opcode: QUERY, rcode: NOERROR, id: 0
;; flags: rd ; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 0
;; QUESTION SECTION:
;; perf-test-b. IN A
;; ANSWER SECTION:
;; AUTHORITY SECTION:
;; ADDITIONAL SECTION:
;; Query time: 0 msec
;; WHEN: Thu Aug 18 02:08:47 2016
;; MSG SIZE rcvd: 0
;; ->>HEADER<<- opcode: QUERY, rcode: NOERROR, id: 52723
;; flags: qr rd ra ; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0
;; QUESTION SECTION:
;; perf-test-b. IN A
;; ANSWER SECTION:
perf-test-b. 600 IN A 10.0.3.4 <<<<<<<<<<<<<<<<<<<<<<<<<< Service VIP
;; AUTHORITY SECTION:
;; ADDITIONAL SECTION:
;; Query time: 1 msec
;; SERVER: 127.0.0.11 <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Local resolver
;; WHEN: Thu Aug 18 02:08:47 2016
;; MSG SIZE rcvd: 56
Purpose: a simple Unix utility that reads and writes data across network connections, using the TCP or UDP protocol. It's useful for testing and troubleshooting TCP/UDP connections. If there's a firewall rule blocking certain ports, netcat can be used to detect
🐳 → docker network create -d overlay my-ovl
55rohpeerwqx8og4n0byr0ehu
🐳 → docker service create --name service-a --network my-ovl -p 8080:8080 nicolaka/netshoot nc -l 8080
bnj517hh4ylpf7ewawsp9unrc
🐳 → docker service create --name service-b --network my-ovl nicolaka/netshoot nc -vz service-a 8080
3xv1ukbd3kr03j4uybmmlp27j
🐳 → docker logs service-b.1.0c5wy4104aosovtl1z9oixiso
Connection to service-a 8080 port [tcp/http-alt] succeeded!
purpose: a collection of utilities for controlling TCP / IP networking and traffic control in Linux.
# Sample routing and arp table of the docker host.
🐳 → docker run -it --net host nicolaka/netshoot
/ # ip route show
default via 192.168.65.1 dev eth0 metric 204
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
172.19.0.0/16 dev br-fd694678f5c3 proto kernel scope link src 172.19.0.1 linkdown
172.20.0.0/16 dev docker_gwbridge proto kernel scope link src 172.20.0.1
172.21.0.0/16 dev br-0d73cc4ac114 proto kernel scope link src 172.21.0.1 linkdown
172.22.0.0/16 dev br-1eb1f1e84df8 proto kernel scope link src 172.22.0.1 linkdown
172.23.0.0/16 dev br-aafed4ec941f proto kernel scope link src 172.23.0.1 linkdown
192.168.65.0/29 dev eth0 proto kernel scope link src 192.168.65.2
/ # ip neigh show
192.168.65.1 dev eth0 lladdr f6:16:36:bc:f9:c6 STALE
172.17.0.7 dev docker0 lladdr 02:42:ac:11:00:07 STALE
172.17.0.6 dev docker0 lladdr 02:42:ac:11:00:06 STALE
172.17.0.5 dev docker0 lladdr 02:42:ac:11:00:05 STALE
More info on iproute2 here
Prupose: nsenter is a powerful tool allowing you to enter into any namespaces. nsenter is available inside netshoot but requires netshoot to be run as a privileged container. Additionally, you may want to mount the /var/run/docker/netns directory to be able to enter any network namespace including bridge and overlay networks.
With docker run --name container-B --net container:container-A , docker uses container-A's network namespace ( including interfaces and routes) when creating container-B. This approach is helpful for troubleshooting network issues at the container level. To troubleshoot network issues at the bridge or overlay network level, you need to enter the namespace of the network itself. nsenter allows you to do that.
For example, if we wanted to check the L2 forwarding table for a overlay network. We need to enter the overlay network namespace and use same tools in netshoot to check these entries. The following examples go over some use cases for using nsenter to understand what's happening within a docker network ( overlay in this case).
# Creating an overlay network
🐳 → docker network create -d overlay nsenter-test
9tp0f348donsdj75pktssd97b
# Launching a simple busybox service with 3 replicas
🐳 → docker service create --name nsenter-l2-table-test --replicas 3 --network nsenter-test busybox ping localhost
3692i3q3u8nephdco2c10ro4c
# Inspecting the service
🐳 → docker network inspect nsenter-test
[
{
"Name": "nsenter-test",
"Id": "9tp0f348donsdj75pktssd97b",
"Scope": "swarm",
"Driver": "overlay",
"EnableIPv6": false,
"IPAM": {
"Driver": "default",
"Options": null,
"Config": [
{
"Subnet": "10.0.1.0/24",
"Gateway": "10.0.1.1"
}
]
},
"Internal": false,
"Containers": {
"0ebe0fab555d2e2ef2fcda634bef2071ad3f5842b06bd134b40f259ab9be4f13": {
"Name": "nsenter-l2-table-test.2.83uezc16jcaz2rp6cjwyf4605",
"EndpointID": "3064946bb0224a4b3647cefcba18dcbea71b90a2ba1c09212a7bc599ec1ed3eb",
"MacAddress": "02:42:0a:00:01:04",
"IPv4Address": "10.0.1.4/24",
"IPv6Address": ""
},
"55065360ac1c71638fdef50a073a661dec53b693409c5e09f8f854abc7dbb373": {
"Name": "nsenter-l2-table-test.1.4ryh3wmmv21nsrfwmilanypqq",
"EndpointID": "f81ae5f979d6c54f60636ca9bb2107d95ebf9a08f64786c549e87a66190f1b1f",
"MacAddress": "02:42:0a:00:01:03",
"IPv4Address": "10.0.1.3/24",
"IPv6Address": ""
},
"57eca277749bb01a488f0e6c4e91dc6720b7c8f08531536377b29a972971f54b": {
"Name": "nsenter-l2-table-test.3.9cuoq5m2ue1wi4lsw64k88tvz",
"EndpointID": "ff1a251ffd6c674cd5fd117386d1a197ab68b4ed708187035d91ff5bd5fe0251",
"MacAddress": "02:42:0a:00:01:05",
"IPv4Address": "10.0.1.5/24",
"IPv6Address": ""
}
},
"Options": {
"com.docker.network.driver.overlay.vxlanid_list": "260"
},
"Labels": {}
}
]
# Launcing netshoot in privileged mode
🐳 → docker run -it --rm -v /var/run/docker/netns:/var/run/docker/netns --privileged=true nicolaka/netshoot
# Listing all docker-created network namespaces
/ # cd /var/run/docker/netns/
/var/run/docker/netns # ls
0b1b36d33313 1-9tp0f348do 14d1428c3962 645eb414b538 816b96054426 916dbaa7ea76 db9fd2d68a9b e79049ce9994 f857b5c01ced
1-9r17dodsxt 1159c401b8d8 1a508036acc8 7ca29d89293c 83b743f2f087 aeed676a57a5 default f22ffa5115a0
# The overlay network that we created had an id of 9tp0f348donsdj75pktssd97b. All overlay networks are named <number>-<id>. We can see it in the list as `1-9tp0f348do`. To enter it:
/ # nsenter --net=/var/run/docker/netns/1-9tp0f348do sh
# Now all the commands we issue are within that namespace.
/ # ifconfig
br0 Link encap:Ethernet HWaddr 02:15:B8:E7:DE:B3
inet addr:10.0.1.1 Bcast:0.0.0.0 Mask:255.255.255.0
inet6 addr: fe80::20ce:a5ff:fe63:437d%32621/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:36 errors:0 dropped:0 overruns:0 frame:0
TX packets:18 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:2224 (2.1 KiB) TX bytes:1348 (1.3 KiB)
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1%32621/128 Scope:Host
UP LOOPBACK RUNNING MTU:65536 Metric:1
RX packets:4 errors:0 dropped:0 overruns:0 frame:0
TX packets:4 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1
RX bytes:336 (336.0 B) TX bytes:336 (336.0 B)
veth2 Link encap:Ethernet HWaddr 02:15:B8:E7:DE:B3
inet6 addr: fe80::15:b8ff:fee7:deb3%32621/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:9 errors:0 dropped:0 overruns:0 frame:0
TX packets:32 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:690 (690.0 B) TX bytes:2460 (2.4 KiB)
veth3 Link encap:Ethernet HWaddr 7E:55:C3:5C:C2:78
inet6 addr: fe80::7c55:c3ff:fe5c:c278%32621/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:13 errors:0 dropped:0 overruns:0 frame:0
TX packets:26 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:970 (970.0 B) TX bytes:1940 (1.8 KiB)
veth4 Link encap:Ethernet HWaddr 72:95:AB:A1:6A:87
inet6 addr: fe80::7095:abff:fea1:6a87%32621/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:14 errors:0 dropped:0 overruns:0 frame:0
TX packets:27 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:1068 (1.0 KiB) TX bytes:2038 (1.9 KiB)
vxlan1 Link encap:Ethernet HWaddr EA:EC:1D:B1:7D:D7
inet6 addr: fe80::e8ec:1dff:feb1:7dd7%32621/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:0 errors:0 dropped:0 overruns:0 frame:0
TX packets:0 errors:0 dropped:33 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:0 (0.0 B) TX bytes:0 (0.0 B)
# Let's check out the L2 forwarding table. These MAC addresses belong to the tasks/containers in this service.
/ # bridge fdb show br br0
33:33:00:00:00:01 dev br0 self permanent
01:00:5e:00:00:01 dev br0 self permanent
33:33:ff:63:43:7d dev br0 self permanent
ea:ec:1d:b1:7d:d7 dev vxlan1 master br0 permanent
02:15:b8:e7:de:b3 dev veth2 master br0 permanent
33:33:00:00:00:01 dev veth2 self permanent
01:00:5e:00:00:01 dev veth2 self permanent
33:33:ff:e7:de:b3 dev veth2 self permanent
7e:55:c3:5c:c2:78 dev veth3 master br0 permanent
33:33:00:00:00:01 dev veth3 self permanent
01:00:5e:00:00:01 dev veth3 self permanent
33:33:ff:5c:c2:78 dev veth3 self permanent
72:95:ab:a1:6a:87 dev veth4 master br0 permanent
33:33:00:00:00:01 dev veth4 self permanent
01:00:5e:00:00:01 dev veth4 self permanent
33:33:ff:a1:6a:87 dev veth4 self permanent
# ARP and routing tables. Note that an overlay network only routes traffic for that network. It only has a single route that matches the subnet of that network.
/ # ip neigh show
/ # ip route
10.0.1.0/24 dev br0 proto kernel scope link src 10.0.1.1
# Looks like the arp table is flushed. Let's ping some of the containers on this network.
/ # ping 10.0.1.4
PING 10.0.1.4 (10.0.1.4) 56(84) bytes of data.
64 bytes from 10.0.1.4: icmp_seq=1 ttl=64 time=0.207 ms
64 bytes from 10.0.1.4: icmp_seq=2 ttl=64 time=0.087 ms
^C
--- 10.0.1.4 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1002ms
rtt min/avg/max/mdev = 0.087/0.147/0.207/0.060 ms
/ # ip neigh show
10.0.1.4 dev br0 lladdr 02:42:0a:00:01:04 REACHABLE
# and using bridge-utils to show interfaces of the overlay network local bridge.
/ # brctl show
bridge name bridge id STP enabled interfaces
br0 8000.0215b8e7deb3 no vxlan1
veth2
veth3
veth4
Feel free to provide feedback and contribute networking troubleshooting tools and use-cases by opening PRs.