Building of Self-Installer package from source code

OOM ONAP offline installer
Self-Installer Package Build Guide
(Community Edition)

This document is describing procedure for generating SI (self-installer) archive based on source code.

It's supposed to be triggered on server with internet connectivity and will download all artifacts required for stable/beijing deployment based on our static lists.
Procedure was completely tested on RHEL 7.4 as it's our target platform, however with small adaptations it should be applicable also for other platforms.

Part 1. Preparations
Part 2. Download artifacts for offline installer
Part 3. Populate local nexus
Part 4. Creating self-install archive:

Version Control

Version	Date	Modified by	Comment
0.1	13.10.2018	Michal Ptacek	First draft

Contributors

Name	Mail
Michal Ptá?ek	m.ptacek@partner.samsung.com
Samuli Silvius	s.silvius@partner.samsung.com
Timo Puha	t.puha@partner.samsung.com
Petr Ospalý	p.ospaly@partner.samsung.com
Pawel Mentel	p.mentel@samsung.com
Witold Kopeld	w.kopel@samsung.com

Part 1. Preparations

We assume that procedure is executed on:

Rhel 7.4 server with ~200G disc space, 16G+ RAM
Internet connectivity

More-over following packages has to be installed

screen expect nodejs git wget createrepo python2-pip patch
docker (exact version, from centos repo)
source code of offline installer

This can be achieved by following commands:

Register server (rhel only)

 subscription-manager register --username <rhel licence name> --password <password> --auto-attach

enable EPEL

yum install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm
# enable rhel-7-server-e4s-optional-rpms in /etc/yum.repos.d/redhat.repo
# to get following rpm available (icewm dependency)
# fribidi x86_64 0.19.4-6.el7 rhel-7-server-e4s-optional-rpms

install following packages

yum install -y screen expect nodejs git wget createrepo python2-pip patch yum-utils

install docker

curl https://releases.rancher.com/install-docker/17.03.sh | sh

install docker alternative - both rhel and centos:

yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
yum install docker-ce-17.03.3.ce-1.el7 docker-ce-selinux-17.03.3.ce-1.el7

start docker
```
systemctl start docker
```

7. download the installer

git clone  https://git.onap.org/integration/devtool
cd devtool
git checkout remotes/origin/beijing
cd onap-offline

Part 2. Download artifacts for offline installer

All artifacts should be downloaded by running following script

 ./bash/tools/download_offline_data_by_lists.sh

Download is as reliable as network connectivity to internet, it's highly recommended to run it in screen and save log file from this script execution for checking

if all artifacts were successfully collected. Each start and end of script call should contain timestamp in console output.

Downloading consists of 12 steps, which should be checked at the end one-by-one.

*Verify:* +Please take a look on following comments to respective parts of download script+

[Step 1/12 Download collected docker images]

[Step 2/12 Download manually collected docker images]

both image download steps are quite reliable and contain retry logic

E.g

== pkg #143 of 163 ==
rancher/etc-host-updater:v0.0.3 digest:sha256:bc156a5ae480d6d6d536aa454a9cc2a88385988617a388808b271e06dc309ce8
Error response from daemon: Get https://registry-1.docker.io/v2/rancher/etc-host-updater/manifests/v0.0.3: Get https://auth.docker.io/token?scope=repository%3Arancher%2Fetc-host-updater%3Apull&service=registry.docker.io: net/http: TLS handshake timeout
WARNING [!]: warning Command docker -l error pull rancher/etc-host-updater:v0.0.3 failed. Attempt: 2/5
INFO: info waiting 10s for another try...
v0.0.3: Pulling from rancher/etc-host-updater
b3e1c725a85f: Already exists
6a710864a9fc: Already exists
d0ac3b234321: Already exists
87f567b5cf58: Already exists
16914729cfd3: Already exists
83c2da5790af: Pulling fs layer
83c2da5790af: Verifying Checksum
83c2da5790af: Download complete
83c2da5790af: Pull complete

[Step 3/12 Build own nginx image]

there is no hardening in this step, if it failed needs to be retriggered. It should end with “Successfully built <id>”

[Step 4/12 Save docker images from docker cache to tarfiles]

quite reliable, retry logic in place

[Step 5/12 move infra related images to infra folder]

should be safe, precondition is not failing step(3)

[Step 6/12 Download git repos]

potentially unsafe, no hardening in place. If it not download all git repos. It has to be executed again. Easiest way is probably to comment-out other steps in load script and run it again.

E.g.

Cloning into bare repository 'github.com/rancher/community-catalog.git'...
error: RPC failed; result=28, HTTP code = 0
fatal: The remote end hung up unexpectedly
Cloning into bare repository 'git.rancher.io/rancher-catalog.git'...
Cloning into bare repository 'gerrit.onap.org/r/testsuite/properties.git'...
Cloning into bare repository 'gerrit.onap.org/r/portal.git'...
Cloning into bare repository 'gerrit.onap.org/r/aaf/authz.git'...
Cloning into bare repository 'gerrit.onap.org/r/demo.git'...
Cloning into bare repository 'gerrit.onap.org/r/dmaap/messagerouter/messageservice.git'...
Cloning into bare repository 'gerrit.onap.org/r/so/docker-config.git'...

[Step 7/12 Download http files]

[Step 8/12 Download npm pkgs]

[Step 9/12 Download bin tools]

work quite reliably, If it not download all artifacts. Easiest way is probably to comment-out other steps in load script and run it again.

[Step 10/12 Download rhel pkgs]

this is the step which will work on rhel only, for other platform different packages has to be downloaded.

We need just couple of rpms, but those has a lot of dependencies (mostly because of vnc).

Script is also download all perl packages from all repos, but we need around dozen of them.

Following is considered as successful run of this part:

    Available: 1:net-snmp-devel-5.7.2-32.el7.i686 (rhel-7-server-rpms)
       net-snmp-devel = 1:5.7.2-32.el7
   Available: 1:net-snmp-devel-5.7.2-33.el7_5.2.i686 (rhel-7-server-rpms)
       net-snmp-devel = 1:5.7.2-33.el7_5.2
Dependency resolution failed, some packages will not be downloaded.
No Presto metadata available for rhel-7-server-rpms
https://ftp.icm.edu.pl/pub/Linux/fedora/linux/epel/7/x86_64/Packages/p/perl-CDB_File-0.98-9.el7.x86_64.rpm: [Errno 12] Timeout on https://ftp.icm.edu.pl/pub/Linux/fedora/linux/epel/7/x86_64/Packages/p/perl-CDB_File-0.98-9.el7.x86_64.rpm:
(28, 'Operation timed out after 30001 milliseconds with 0 out of 0 bytes received')
Trying other mirror.
Spawning worker 0 with 230 pkgs
Spawning worker 1 with 230 pkgs
Spawning worker 2 with 230 pkgs
Spawning worker 3 with 230 pkgs
Spawning worker 4 with 229 pkgs
Spawning worker 5 with 229 pkgs
Spawning worker 6 with 229 pkgs
Spawning worker 7 with 229 pkgs
Workers Finished
Saving Primary metadata
Saving file lists metadata
Saving other metadata
Generating sqlite DBs
Sqlite DBs complete

[Step 11/12 Download oom]

this step is downloading oom repo into ./resources/oom and patch it using our patch file. If this step is retried after previously passing it will lead to inconsistent oom repo.

Because patch will fail and create some .rej files, which will be marked as broken during onap_deploy part later.

E.g. successful run looks like this:

Checkout base commit which will be patched
Switched to a new branch 'patched_beijing'
patching file kubernetes/appc/values.yaml
patching file kubernetes/common/dgbuilder/templates/deployment.yaml
patching file kubernetes/dcaegen2/charts/dcae-cloudify-manager/templates/deployment.yaml
patching file kubernetes/dmaap/charts/message-router/templates/deployment.yaml
patching file kubernetes/onap/values.yaml
patching file kubernetes/policy/charts/drools/resources/config/opt/policy/config/drools/apps-install.sh
patching file kubernetes/policy/charts/drools/resources/scripts/update-vfw-op-policy.sh
patching file kubernetes/policy/resources/config/pe/push-policies.sh
patching file kubernetes/robot/values.yaml
patching file kubernetes/sdnc/charts/sdnc-ansible-server/templates/deployment.yaml
patching file kubernetes/sdnc/charts/sdnc-portal/templates/deployment.yaml
patching file kubernetes/uui/charts/uui-server/templates/deployment.yaml

[Step 12/12 Download sdnc-ansible-server packages]

there is again no retry logic in this part, it’s collecting packages for sdnc-ansible-server in the exactly same way how that container is doing it,

however there is a bug in upstream that image in place won’t work with those packages as old ones are not available and newer are not compatible with other stuff inside that image

Following is approximate size of all artifacts after successful download

[root@upstream-master onap-offline]# for i in `ls -1 resources/`;do du -h resources/$i | tail -1;done
126M    resources/downloads
97M     resources/git-repo
61M     resources/http
91G     resources/offline_data
36M     resources/oom
638M    resources/pkg

Part 3. Populate local nexus

Prereq:

Cleaned repos (ensure that we use our packages from onap.repo only)

mkdir /etc/yum.repos.d-backup && mv /etc/yum.repos.d/*.repo /etc/yum.repos.d-backup/

actual directory with local_repo.conf
```
./bash/tools/deploy_nexus.sh 
```
User is prompted for local ip, where nexus should be deployed
For accessing nexus GUI, feel free to use preconfigured VNC server (adjust the firewall on the host server)
One just need to create ssh tunnel to vcnserver and connect it with vncviewer
vncserver listens on 5901 and password is onap
```
$ ssh -NfL 1234:127.0.0.1:5901 <the host server>
$ vncviewer 127.0.0.1::1234
```
Nexus must be accessed from one of nginx simulated domains,
e.g. http://nexus3.onap.org (Because on port 80 is more services/domains, and it is distinguished by his fqdn on proxy)

VNCserver is just example how to access nexus gui, alternatively tunnel can be created. All of this is not needed if server can be reached directly.

Following stuff should be configured from nexus gui:

Login (default: admin / admin123)
Settings -> repositories > create repository #docker
- Select docker (hosted)
- Name: anything you want (ex: onap)
- Repository connectors: HTTP set to 8082
- Force basic authentication: UNchecked
- Enable docker V1 API: checked
Settings -> Repositories > create repository #npm
- Select npm (hosted)
- Name: npm-private
- Deployment policy: Allow redeploy
Settings -> Repositories -> create repository #maven
- Select maven2 (hosted)
- Name: maven2
- Deployment policy: Allow redeploy
Settings -> Security -> Realms
- Add Docker Bearer Token Realm
- Add npm Bearer Token Realm
Settings -> Security -> Users
- Add user docker / pwd docker
- Rights same as anonymous

docker

#Once nexus is created and configured, use following script

to load docker / npm & maven artifacts into it
```
./bash/tools/load_stored_offline_data.sh
```
Once all artifacts are safely loaded into nexus, original directory with tar files of images

can be removed, we don't need them inside SI archive anymore
E.g

rm -rf ./resources/offline_data/docker_images_for_nexus/*

Part 4. Creating self-install archive:

Note: as nexus_data is mounted folder into nexus, we recommend to stop nexus first before launching SI archive create script

E.g.

[root@upstream-master onap-offline]# docker ps
CONTAINER ID        IMAGE COMMAND                  CREATED STATUS PORTS                                                              NAMES
74d50a5d5212        own_nginx "/bin/sh -c 'spawn..."   7 hours ago Up 7 hours 0.0.0.0:80->80/tcp, 0.0.0.0:443->443/tcp, 0.0.0.0:10001->443/tcp   nginx
b73ae7b76d71        sonatype/nexus3 "sh -c ${SONATYPE_..."   7 hours ago Up 7 hours 8081/tcp                                                           nexus

[root@upstream-master onap-offline]# docker stop b73ae7b76d71
b73ae7b76d71

SI (self-installer) package can be created with prepopulated data via following command

./bash/tools/create_si_onap_pkg.sh <package_suffix_name>

E.g. ./bash/tools/create_si_onap_pkg.sh beijing1