Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 1 | Simulating networks with VPP |
| 2 | ============================ |
| 3 | |
| 4 | The “make test” framework provides a good way to test individual |
| 5 | features. However, when testing several features at once - or validating |
| 6 | nontrivial configurations - it may prove difficult or impossible to use |
| 7 | the unit-test framework. |
| 8 | |
| 9 | This note explains how to set up lxc/lxd, and a 5-container testbed to |
| 10 | test a split-tunnel nat + ikev2 + ipsec + ipv6 prefix-delegation |
| 11 | scenario. |
| 12 | |
| 13 | OS / Distro test results |
| 14 | ------------------------ |
| 15 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 16 | This setup has been tested on Ubuntu 18.04 - 22.04 LTS systems. Other |
| 17 | distros may work fine, or not at all. |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 18 | |
| 19 | Proxy Server |
| 20 | ------------ |
| 21 | |
| 22 | If you need to use a proxy server e.g. from a lab system, you’ll |
| 23 | probably need to set HTTP_PROXY, HTTPS_PROXY, http_proxy and https_proxy |
| 24 | in /etc/environment. Directly setting variables in the environment |
| 25 | doesn’t work. The lxd snap *daemon* needs the proxy settings, not the |
| 26 | user interface. |
| 27 | |
| 28 | Something like so: |
| 29 | |
| 30 | :: |
| 31 | |
| 32 | HTTP_PROXY=http://my.proxy.server:8080 |
| 33 | HTTPS_PROXY=http://my.proxy.server:4333 |
| 34 | http_proxy=http://my.proxy.server:8080 |
| 35 | https_proxy=http://my.proxy.server:4333 |
| 36 | |
| 37 | Install and configure lxd |
| 38 | ------------------------- |
| 39 | |
| 40 | Install the lxd snap. The lxd snap is up to date, as opposed to the |
| 41 | results of “sudo apt-get install lxd”. |
| 42 | |
| 43 | :: |
| 44 | |
| 45 | # snap install lxd |
| 46 | # lxd init |
| 47 | |
| 48 | “lxd init” asks several questions. With the exception of the storage |
| 49 | pool, take the defaults. To match the configs shown below, create a |
| 50 | storage pool named “vpp.” Storage pools of type “zfs” and “files” have |
| 51 | been tested successfully. |
| 52 | |
| 53 | zfs is more space-efficient. “lxc copy” is infinitely faster with zfs. |
| 54 | The path for the zfs storage pool is under /var. Do not replace it with |
| 55 | a symbolic link, unless you want to rebuild all of your containers from |
| 56 | scratch. Ask me how I know that. |
| 57 | |
| 58 | Create three network segments |
| 59 | ----------------------------- |
| 60 | |
| 61 | Aka, linux bridges. |
| 62 | |
| 63 | :: |
| 64 | |
| 65 | # lxc network create respond |
| 66 | # lxc network create internet |
| 67 | # lxc network create initiate |
| 68 | |
| 69 | We’ll explain the test topology in a bit. Stay tuned. |
| 70 | |
| 71 | Set up the default container profile |
| 72 | ------------------------------------ |
| 73 | |
| 74 | Execute “lxc profile edit default”, and install the following |
| 75 | configuration. Note that the “shared” directory should mount your vpp |
| 76 | workspaces. With that trick, you can edit code from any of the |
| 77 | containers, run vpp without installing it, etc. |
| 78 | |
| 79 | :: |
| 80 | |
| 81 | config: {} |
| 82 | description: Default LXD profile |
| 83 | devices: |
| 84 | eth0: |
| 85 | name: eth0 |
| 86 | network: lxdbr0 |
| 87 | type: nic |
| 88 | eth1: |
| 89 | name: eth1 |
| 90 | nictype: bridged |
| 91 | parent: internet |
| 92 | type: nic |
| 93 | eth2: |
| 94 | name: eth2 |
| 95 | nictype: bridged |
| 96 | parent: respond |
| 97 | type: nic |
| 98 | eth3: |
| 99 | name: eth3 |
| 100 | nictype: bridged |
| 101 | parent: initiate |
| 102 | type: nic |
| 103 | root: |
| 104 | path: / |
| 105 | pool: vpp |
| 106 | type: disk |
| 107 | shared: |
| 108 | path: /scratch |
| 109 | source: /scratch |
| 110 | type: disk |
| 111 | name: default |
| 112 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 113 | |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 114 | Set up the network configurations |
| 115 | --------------------------------- |
| 116 | |
| 117 | Edit the fake “internet” backbone: |
| 118 | |
| 119 | :: |
| 120 | |
| 121 | # lxc network edit internet |
| 122 | |
| 123 | Install the ip addresses shown below, to avoid having to rebuild the vpp |
| 124 | and host configuration: |
| 125 | |
| 126 | :: |
| 127 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 128 | config: |
| 129 | ipv4.address: 10.26.68.1/24 |
| 130 | ipv4.dhcp.ranges: 10.26.68.10-10.26.68.50 |
| 131 | ipv4.nat: "true" |
| 132 | ipv6.address: none |
| 133 | ipv6.nat: "false" |
| 134 | description: "" |
| 135 | name: internet |
| 136 | type: bridge |
| 137 | used_by: |
| 138 | managed: true |
| 139 | status: Created |
| 140 | locations: |
| 141 | - none |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 142 | |
| 143 | Repeat the process with the “respond” and “initiate” networks, using |
| 144 | these configurations: |
| 145 | |
| 146 | respond network configuration |
| 147 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 148 | |
| 149 | :: |
| 150 | |
| 151 | config: |
| 152 | ipv4.address: 10.166.14.1/24 |
| 153 | ipv4.dhcp.ranges: 10.166.14.10-10.166.14.50 |
| 154 | ipv4.nat: "true" |
| 155 | ipv6.address: none |
| 156 | ipv6.nat: "false" |
| 157 | description: "" |
| 158 | name: respond |
| 159 | type: bridge |
| 160 | used_by: |
| 161 | managed: true |
| 162 | status: Created |
| 163 | locations: |
| 164 | - none |
| 165 | |
| 166 | initiate network configuration |
| 167 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 168 | |
| 169 | :: |
| 170 | |
| 171 | config: |
| 172 | ipv4.address: 10.219.188.1/24 |
| 173 | ipv4.dhcp.ranges: 10.219.188.10-10.219.188.50 |
| 174 | ipv4.nat: "true" |
| 175 | ipv6.address: none |
| 176 | ipv6.nat: "false" |
| 177 | description: "" |
| 178 | name: initiate |
| 179 | type: bridge |
| 180 | used_by: |
| 181 | managed: true |
| 182 | status: Created |
| 183 | locations: |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 184 | |
| 185 | Create a “master” container image |
| 186 | --------------------------------- |
| 187 | |
| 188 | The master container image should be set up so that you can build vpp, |
| 189 | ssh into the container, edit source code, run gdb, etc. |
| 190 | |
| 191 | Make sure that e.g. public key auth ssh works. |
| 192 | |
| 193 | :: |
| 194 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 195 | # lxd launch ubuntu:22.04 respond |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 196 | <spew> |
| 197 | # lxc exec respond bash |
| 198 | respond# cd /scratch/my-vpp-workspace |
| 199 | respond# apt-get install make ssh |
| 200 | respond# make install-dep |
| 201 | respond# exit |
| 202 | # lxc stop respond |
| 203 | |
| 204 | Mark the container image privileged. If you forget this step, you’ll |
| 205 | trip over a netlink error (-11) aka EAGAIN when you try to roll in the |
| 206 | vpp configurations. |
| 207 | |
| 208 | :: |
| 209 | |
| 210 | # lxc config set respond security.privileged "true" |
| 211 | |
| 212 | Duplicate the “master” container image |
| 213 | -------------------------------------- |
| 214 | |
| 215 | To avoid having to configure N containers, be sure that the master |
| 216 | container image is fully set up before you help it have children: |
| 217 | |
| 218 | :: |
| 219 | |
| 220 | # lxc copy respond respondhost |
| 221 | # lxc copy respond initiate |
| 222 | # lxc copy respond initiatehost |
| 223 | # lxc copy respond dhcpserver # optional, to test ipv6 prefix delegation |
| 224 | |
| 225 | Install handy script |
| 226 | -------------------- |
| 227 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 228 | See below for a handy host script which executes lxc commands across |
| 229 | the current set of running containers. I call it “lxc-foreach,” feel |
| 230 | free to call the script Ishmael if you like. |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 231 | |
| 232 | Examples: |
| 233 | |
| 234 | :: |
| 235 | |
| 236 | $ lxc-foreach start |
| 237 | <issues "lxc start" for each container in the list> |
| 238 | |
| 239 | After a few seconds, use this one to open an ssh connection to each |
| 240 | container. The ssh command parses the output of “lxc info,” which |
| 241 | displays container ip addresses. |
| 242 | |
| 243 | :: |
| 244 | |
| 245 | $ lxc-foreach ssh |
| 246 | |
| 247 | Here’s the script: |
| 248 | |
| 249 | :: |
| 250 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 251 | #!/bin/bash |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 252 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 253 | set -u |
| 254 | export containers="respond initiate initiatehost respondhost" |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 255 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 256 | if [ x$1 = "x" ] ; then |
| 257 | echo missing command |
| 258 | exit 1 |
| 259 | fi |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 260 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 261 | if [ $1 = "ssh" ] ; then |
| 262 | for c in $containers |
| 263 | do |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 264 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 265 | inet=`lxc info $c | grep 10.38.33 | sed "s/.*inet://" | sed "s/\/24.*//" | tr -d " "` |
| 266 | if [ x$inet != "x" ] ; then |
| 267 | gnome-terminal --title "$c(ssh)" --command "/usr/bin/ssh -Y root@$inet" |
| 268 | fi |
| 269 | done |
| 270 | exit 0 |
| 271 | fi |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 272 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 273 | for c in $containers |
| 274 | do |
| 275 | inet=`lxc info $c | grep 10.38.33 | sed "s/.*inet://" | sed "s/\/24.*//" | tr -d " "` |
| 276 | if [ x$1 = "xstart" ] ; then |
| 277 | echo Starting $c |
| 278 | lxc start $c |
| 279 | elif [ x$1 = "xstop" ] ; then |
| 280 | echo Stopping $c |
| 281 | lxc stop $c |
| 282 | elif [ x$inet != "x" ] ; then |
| 283 | echo lxc $1 $c |
| 284 | lxc $1 $c |
| 285 | fi |
| 286 | done |
| 287 | |
| 288 | exit 0 |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 289 | |
| 290 | Test topology |
| 291 | ------------- |
| 292 | |
| 293 | Finally, we’re ready to describe a test topology. First, a picture: |
| 294 | |
| 295 | :: |
| 296 | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 297 | ===+======== management lan/bridge lxdbr0 (dhcp) ==============+=== |
| 298 | | | | |
| 299 | | | | |
| 300 | | | | |
| 301 | v | v |
| 302 | eth0 | eth0 |
| 303 | +---------+ eth1 eth1 +----------+ |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 304 | | respond | 10.26.88.100 <= internet bridge => 10.26.88.101 | initiate | |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 305 | +---------+ +----------+ |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 306 | eth2 / bvi0 10.166.14.2 | 10.219.188.2 eth3 / bvi0 |
Dave Barach | b7e66f4 | 2023-12-29 11:15:34 -0500 | [diff] [blame^] | 307 | | | | |
| 308 | | ("respond" bridge) | ("initiate" bridge) | |
| 309 | | | | |
| 310 | v | v |
| 311 | eth2 10.166.14.3 | eth3 10.219.188.3 |
| 312 | +-------------+ | +-------------+ |
| 313 | | respondhost | | | respondhost | |
| 314 | +-------------+ | +-------------+ |
| 315 | v |
Nathan Skrzypczak | 9ad39c0 | 2021-08-19 11:38:06 +0200 | [diff] [blame] | 316 | eth0 (management lan) <========+========> eth0 (management lan) |
| 317 | |
| 318 | Test topology discussion |
| 319 | ~~~~~~~~~~~~~~~~~~~~~~~~ |
| 320 | |
| 321 | This topology is suitable for testing almost any tunnel encap/decap |
| 322 | scenario. The two containers “respondhost” and “initiatehost” are |
| 323 | end-stations connected to two vpp instances running on “respond” and |
| 324 | “initiate”. |
| 325 | |
| 326 | We leverage the Linux end-station network stacks to generate traffic of |
| 327 | all sorts. |
| 328 | |
| 329 | The so-called “internet” bridge models the public internet. The |
| 330 | “respond” and “initiate” bridges connect vpp instances to local hosts |
| 331 | |
| 332 | End station configs |
| 333 | ------------------- |
| 334 | |
| 335 | The end-station Linux configurations set up the eth2 and eth3 ip |
| 336 | addresses shown above, and add tunnel routes to the opposite end-station |
| 337 | networks. |
| 338 | |
| 339 | respondhost configuration |
| 340 | ~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 341 | |
| 342 | :: |
| 343 | |
| 344 | ifconfig eth2 10.166.14.3/24 up |
| 345 | route add -net 10.219.188.0/24 gw 10.166.14.2 |
| 346 | |
| 347 | initiatehost configuration |
| 348 | ~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 349 | |
| 350 | :: |
| 351 | |
| 352 | sudo ifconfig eth3 10.219.188.3/24 up |
| 353 | sudo route add -net 10.166.14.0/24 gw 10.219.188.2 |
| 354 | |
| 355 | VPP configs |
| 356 | ----------- |
| 357 | |
| 358 | Split nat44 / ikev2 + ipsec tunneling, with ipv6 prefix delegation in |
| 359 | the “respond” config. |
| 360 | |
| 361 | respond configuration |
| 362 | ~~~~~~~~~~~~~~~~~~~~~ |
| 363 | |
| 364 | :: |
| 365 | |
| 366 | set term pag off |
| 367 | |
| 368 | comment { "internet" } |
| 369 | create host-interface name eth1 |
| 370 | set int ip address host-eth1 10.26.68.100/24 |
| 371 | set int ip6 table host-eth1 0 |
| 372 | set int state host-eth1 up |
| 373 | |
| 374 | comment { default route via initiate } |
| 375 | ip route add 0.0.0.0/0 via 10.26.68.101 |
| 376 | |
| 377 | comment { "respond-private-net" } |
| 378 | create host-interface name eth2 |
| 379 | bvi create instance 0 |
| 380 | set int l2 bridge bvi0 1 bvi |
| 381 | set int ip address bvi0 10.166.14.2/24 |
| 382 | set int state bvi0 up |
| 383 | set int l2 bridge host-eth2 1 |
| 384 | set int state host-eth2 up |
| 385 | |
| 386 | |
| 387 | nat44 add interface address host-eth1 |
| 388 | set interface nat44 in host-eth2 out host-eth1 |
| 389 | nat44 add identity mapping external host-eth1 udp 500 |
| 390 | nat44 add identity mapping external host-eth1 udp 4500 |
| 391 | comment { nat44 untranslated subnet 10.219.188.0/24 } |
| 392 | |
| 393 | comment { responder profile } |
| 394 | ikev2 profile add initiate |
| 395 | ikev2 profile set initiate udp-encap |
| 396 | ikev2 profile set initiate auth rsa-sig cert-file /scratch/setups/respondcert.pem |
| 397 | set ikev2 local key /scratch/setups/initiatekey.pem |
| 398 | ikev2 profile set initiate id local fqdn initiator.my.net |
| 399 | ikev2 profile set initiate id remote fqdn responder.my.net |
| 400 | ikev2 profile set initiate traffic-selector remote ip-range 10.219.188.0 - 10.219.188.255 port-range 0 - 65535 protocol 0 |
| 401 | ikev2 profile set initiate traffic-selector local ip-range 10.166.14.0 - 10.166.14.255 port-range 0 - 65535 protocol 0 |
| 402 | create ipip tunnel src 10.26.68.100 dst 10.26.68.101 |
| 403 | ikev2 profile set initiate tunnel ipip0 |
| 404 | |
| 405 | comment { ipv6 prefix delegation } |
| 406 | ip6 nd address autoconfig host-eth1 default-route |
| 407 | dhcp6 client host-eth1 |
| 408 | dhcp6 pd client host-eth1 prefix group hgw |
| 409 | set ip6 address bvi0 prefix group hgw ::2/56 |
| 410 | ip6 nd address autoconfig bvi0 default-route |
| 411 | ip6 nd bvi0 ra-interval 5 3 ra-lifetime 180 |
| 412 | |
| 413 | set int mtu packet 1390 ipip0 |
| 414 | set int unnum ipip0 use host-eth1 |
| 415 | ip route add 10.219.188.0/24 via ipip0 |
| 416 | |
| 417 | initiate configuration |
| 418 | ~~~~~~~~~~~~~~~~~~~~~~ |
| 419 | |
| 420 | :: |
| 421 | |
| 422 | set term pag off |
| 423 | |
| 424 | comment { "internet" } |
| 425 | create host-interface name eth1 |
| 426 | comment { set dhcp client intfc host-eth1 hostname initiate } |
| 427 | set int ip address host-eth1 10.26.68.101/24 |
| 428 | set int state host-eth1 up |
| 429 | |
| 430 | comment { default route via "internet gateway" } |
| 431 | comment { ip route add 0.0.0.0/0 via 10.26.68.1 } |
| 432 | |
| 433 | comment { "initiate-private-net" } |
| 434 | create host-interface name eth3 |
| 435 | bvi create instance 0 |
| 436 | set int l2 bridge bvi0 1 bvi |
| 437 | set int ip address bvi0 10.219.188.2/24 |
| 438 | set int state bvi0 up |
| 439 | set int l2 bridge host-eth3 1 |
| 440 | set int state host-eth3 up |
| 441 | |
| 442 | nat44 add interface address host-eth1 |
| 443 | set interface nat44 in bvi0 out host-eth1 |
| 444 | nat44 add identity mapping external host-eth1 udp 500 |
| 445 | nat44 add identity mapping external host-eth1 udp 4500 |
| 446 | comment { nat44 untranslated subnet 10.166.14.0/24 } |
| 447 | |
| 448 | comment { initiator profile } |
| 449 | ikev2 profile add respond |
| 450 | ikev2 profile set respond udp-encap |
| 451 | ikev2 profile set respond auth rsa-sig cert-file /scratch/setups/initiatecert.pem |
| 452 | set ikev2 local key /scratch/setups/respondkey.pem |
| 453 | ikev2 profile set respond id local fqdn responder.my.net |
| 454 | ikev2 profile set respond id remote fqdn initiator.my.net |
| 455 | |
| 456 | ikev2 profile set respond traffic-selector remote ip-range 10.166.14.0 - 10.166.14.255 port-range 0 - 65535 protocol 0 |
| 457 | ikev2 profile set respond traffic-selector local ip-range 10.219.188.0 - 10.219.188.255 port-range 0 - 65535 protocol 0 |
| 458 | |
| 459 | ikev2 profile set respond responder host-eth1 10.26.68.100 |
| 460 | ikev2 profile set respond ike-crypto-alg aes-cbc 256 ike-integ-alg sha1-96 ike-dh modp-2048 |
| 461 | ikev2 profile set respond esp-crypto-alg aes-cbc 256 esp-integ-alg sha1-96 esp-dh ecp-256 |
| 462 | ikev2 profile set respond sa-lifetime 3600 10 5 0 |
| 463 | |
| 464 | create ipip tunnel src 10.26.68.101 dst 10.26.68.100 |
| 465 | ikev2 profile set respond tunnel ipip0 |
| 466 | ikev2 initiate sa-init respond |
| 467 | |
| 468 | set int mtu packet 1390 ipip0 |
| 469 | set int unnum ipip0 use host-eth1 |
| 470 | ip route add 10.166.14.0/24 via ipip0 |
| 471 | |
| 472 | IKEv2 certificate setup |
| 473 | ----------------------- |
| 474 | |
| 475 | In both of the vpp configurations, you’ll see “/scratch/setups/xxx.pem” |
| 476 | mentioned. These certificates are used in the ikev2 key exchange. |
| 477 | |
| 478 | Here’s how to generate the certificates: |
| 479 | |
| 480 | :: |
| 481 | |
| 482 | openssl req -x509 -nodes -newkey rsa:4096 -keyout respondkey.pem -out respondcert.pem -days 3560 |
| 483 | openssl x509 -text -noout -in respondcert.pem |
| 484 | openssl req -x509 -nodes -newkey rsa:4096 -keyout initiatekey.pem -out initiatecert.pem -days 3560 |
| 485 | openssl x509 -text -noout -in initiatecert.pem |
| 486 | |
| 487 | Make sure that the “respond” and “initiate” configurations point to the |
| 488 | certificates. |
| 489 | |
| 490 | DHCPv6 server setup |
| 491 | ------------------- |
| 492 | |
| 493 | If you need an ipv6 dhcp server to test ipv6 prefix delegation, create |
| 494 | the “dhcpserver” container as shown above. |
| 495 | |
| 496 | Install the “isc-dhcp-server” Debian package: |
| 497 | |
| 498 | :: |
| 499 | |
| 500 | sudo apt-get install isc-dhcp-server |
| 501 | |
| 502 | /etc/dhcp/dhcpd6.conf |
| 503 | ~~~~~~~~~~~~~~~~~~~~~ |
| 504 | |
| 505 | Edit the dhcpv6 configuration and add an ipv6 subnet with prefix |
| 506 | delegation. For example: |
| 507 | |
| 508 | :: |
| 509 | |
| 510 | subnet6 2001:db01:0:1::/64 { |
| 511 | range6 2001:db01:0:1::1 2001:db01:0:1::9; |
| 512 | prefix6 2001:db01:0:100:: 2001:db01:0:200::/56; |
| 513 | } |
| 514 | |
| 515 | Add an ipv6 address on eth1, which is connected to the “internet” |
| 516 | bridge, and start the dhcp server. I use the following trivial bash |
| 517 | script, which runs the dhcp6 server in the foreground and produces dhcp |
| 518 | traffic spew: |
| 519 | |
| 520 | :: |
| 521 | |
| 522 | #!/bin/bash |
| 523 | ifconfig eth1 inet6 add 2001:db01:0:1::10/64 || true |
| 524 | dhcpd -6 -d -cf /etc/dhcp/dhcpd6.conf |
| 525 | |
| 526 | The “\|\| true” bit keeps going if eth1 already has the indicated ipv6 |
| 527 | address. |
| 528 | |
| 529 | Container / Host Interoperation |
| 530 | ------------------------------- |
| 531 | |
| 532 | Host / container interoperation is highly desirable. If the host and a |
| 533 | set of containers don’t run the same distro *and distro version*, it’s |
| 534 | reasonably likely that the glibc versions won’t match. That, in turn, |
| 535 | makes vpp binaries built in one environment fail in the other. |
| 536 | |
| 537 | Trying to install multiple versions of glibc - especially at the host |
| 538 | level - often ends very badly and is *not recommended*. It’s not just |
| 539 | glibc, either. The dynamic loader ld-linux-xxx-so.2 is glibc version |
| 540 | specific. |
| 541 | |
| 542 | Fortunately, it’s reasonable easy to build lxd container images based on |
| 543 | specific Ubuntu or Debian versions. |
| 544 | |
| 545 | Create a custom root filesystem image |
| 546 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 547 | |
| 548 | First, install the “debootstrap” tool: |
| 549 | |
| 550 | :: |
| 551 | |
| 552 | sudo apt-get install debootstrap |
| 553 | |
| 554 | Make a temp directory, and use debootstrap to populate it. In this |
| 555 | example, we create an Ubuntu 20.04 (focal fossa) base image: |
| 556 | |
| 557 | :: |
| 558 | |
| 559 | # mkdir /tmp/myroot |
| 560 | # debootstrap focal /tmp/myroot http://archive.ubuntu.com/ubuntu |
| 561 | |
| 562 | To tinker with the base image (if desired): |
| 563 | |
| 564 | :: |
| 565 | |
| 566 | # chroot /tmp/myroot |
| 567 | <add packages, etc.> |
| 568 | # exit |
| 569 | |
| 570 | Make a compressed tarball of the base image: |
| 571 | |
| 572 | :: |
| 573 | |
| 574 | # tar zcf /tmp/rootfs.tar.gz -C /tmp/myroot . |
| 575 | |
| 576 | Create a “metadata.yaml” file which describes the base image: |
| 577 | |
| 578 | :: |
| 579 | |
| 580 | architecture: "x86_64" |
| 581 | # To get current date in Unix time, use `date +%s` command |
| 582 | creation_date: 1458040200 |
| 583 | properties: |
| 584 | architecture: "x86_64" |
| 585 | description: "My custom Focal Fossa image" |
| 586 | os: "Ubuntu" |
| 587 | release: "focal" |
| 588 | |
| 589 | Make a compressed tarball of metadata.yaml: |
| 590 | |
| 591 | :: |
| 592 | |
| 593 | # tar zcf metadata.tar.gz metadata.yaml |
| 594 | |
| 595 | Import the image into lxc / lxd: |
| 596 | |
| 597 | :: |
| 598 | |
| 599 | $ lxc image import metadata.tar.gz rootfd.tar.gz --alias focal-base |
| 600 | |
| 601 | Create a container which uses the customized base image: |
| 602 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 603 | |
| 604 | :: |
| 605 | |
| 606 | $ lxc launch focal-base focaltest |
| 607 | $ lxc exec focaltest bash |
| 608 | |
| 609 | The next several steps should be executed in the container, in the bash |
| 610 | shell spun up by “lxc exec…” |
| 611 | |
| 612 | Configure container networking |
| 613 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 614 | |
| 615 | In the container, create /etc/netplan/50-cloud-init.yaml: |
| 616 | |
| 617 | :: |
| 618 | |
| 619 | network: |
| 620 | version: 2 |
| 621 | ethernets: |
| 622 | eth0: |
| 623 | dhcp4: true |
| 624 | |
| 625 | Use “cat > /etc/netplan/50-cloud-init.yaml”, and cut-’n-paste if your |
| 626 | favorite text editor is AWOL. |
| 627 | |
| 628 | Apply the configuration: |
| 629 | |
| 630 | :: |
| 631 | |
| 632 | # netplan apply |
| 633 | |
| 634 | At this point, eth0 should have an ip address, and you should see a |
| 635 | default route with “route -n”. |
| 636 | |
| 637 | Configure apt |
| 638 | ~~~~~~~~~~~~~ |
| 639 | |
| 640 | Again, in the container, set up /etc/apt/sources.list via cut-’n-paste |
| 641 | from a recently update “focal fossa” host. Something like so: |
| 642 | |
| 643 | :: |
| 644 | |
| 645 | deb http://us.archive.ubuntu.com/ubuntu/ focal main restricted |
| 646 | deb http://us.archive.ubuntu.com/ubuntu/ focal-updates main restricted |
| 647 | deb http://us.archive.ubuntu.com/ubuntu/ focal universe |
| 648 | deb http://us.archive.ubuntu.com/ubuntu/ focal-updates universe |
| 649 | deb http://us.archive.ubuntu.com/ubuntu/ focal multiverse |
| 650 | deb http://us.archive.ubuntu.com/ubuntu/ focal-updates multiverse |
| 651 | deb http://us.archive.ubuntu.com/ubuntu/ focal-backports main restricted universe multiverse |
| 652 | deb http://security.ubuntu.com/ubuntu focal-security main restricted |
| 653 | deb http://security.ubuntu.com/ubuntu focal-security universe |
| 654 | deb http://security.ubuntu.com/ubuntu focal-security multiverse |
| 655 | |
| 656 | “apt-get update” and “apt-install” should produce reasonable results. |
| 657 | Suggest “apt-get install make git”. |
| 658 | |
| 659 | At this point, you can use the “/scratch” sharepoint (or similar) to |
| 660 | execute “make install-dep install-ext-deps” to set up the container with |
| 661 | the vpp toolchain; proceed as desired. |