Home

Awesome

Jenkins-based CI for the Scala project

Used for

Old documentation in doc directory; some pieces are still relevant, some are outdated.

History

The AWS infra was originally deployed and maintanied using ansible (this repo). We no longer use ansible to update the infra, but manage it manually (AWS console, ssh to the machines, jenkins UI, artifactory UI).

We are gradually moving away from using this infra

But we still need it because

Webhooks, tokens, accounts etc

SSH access

<details> <summary>Add the following to your <code>~/.ssh/config</code></summary> 
Host jenkins-master
  HostName 54.67.111.226
  User admin

Host jenkins-worker-behemoth-1
  HostName 54.153.2.9
  User admin

Host jenkins-worker-behemoth-2
  HostName 54.153.1.99
  User admin

Host jenkins-worker-behemoth-3
  HostName 54.183.156.89
  User admin

# no public ip, jumphost through master
Host influxdb
  HostName 172.31.0.100
  User ubuntu
  ProxyCommand ssh -q -W %h:%p jenkins-master
</details>

Installed services

nginx

Nginx for scala-ci.typesafe.com on jenkins-master, /etc/nginx/conf.d/jenkins.conf.

Handles jenkins, /artifactory, /grafana, /benchq, ...

Jenkins

Auth goes via GitHub. In the configuraton there's a field "Admin User Names".

AWS Workers are started / stopped by custom jenkins plugin: https://github.com/lightbend-labs/ec2-start-stop

TODO: how to get logs?

Artifactory

scala-ci.typesafe.com/artifactory/ to log in to the UI.

Repositories:

The config file is /opt/jfrog/artifactory/var/etc/system.yaml.

/opt/jfrog/artifactory/var/log/console.log has aggregated logs, logs for individual services in the same directory.

/opt/jfrog/artifactory/var/data/derby is the main database for our artifactory; its large (19G in Aug 2024).

The access service has its own db at /opt/jfrog/artifactory/var/data/access/derby.

Scabot

Scabot triggers Jenkins builds and updates their state on github commits / PRs.

Details

AWS CLI

Disk usage

Use ncdu -x /path to analyze disk usage.

Jenkins workers disk space

Artifactory disk usage

Artifactory disk usage report: https://scala-ci.typesafe.com/ui/admin/monitoring/storage-summary.

<details> <summary>Steps to delete old builds from <code>scala-pr-validation-snapshots</code>:</summary>

Create a file search.json, adjust the cutoff date on the last line:

items.find({
  "repo": "scala-pr-validation-snapshots",
  "$or": [ { "name": { "$match": "scala-compiler*" } }, {"name": { "$match": "scala-reflect*" } }, { "name": { "$match": "scala-library*" } }, { "name": { "$match": "scala-dist*" } }, { "name": { "$match": "scala-partest*" } }, { "name": { "$match": "scalap*" } } ],
  "created": { "$lt": "2020-01-01" }
})

curl -u 'lukas:SEEEKREET' -X POST "https://scala-ci.typesafe.com/artifactory/api/search/aql" -T search.json > artifacts.json

In an up-to-date Scala 2.13.x checkout, the following tests which of the artifacts correspond to revisions that were actually merged into scala/scala. Builds for those revisions are kept, builds for revisions that never made it are added to to-delete.txt.

n=$(cat artifacts.json | jq -r '.results[] | .path' | uniq | wc -l)
for p in $(cat artifacts.json | jq -r '.results[] | .path' | uniq); do
  n=$((n-1))
  sha=$(echo $p | awk -F'-' '{print $(NF-1)}')
  if git branch --contains $sha | grep 2.13.x > /dev/null; then
    echo "$sha y - $n"
  else
    echo "$sha n - $n"
    echo $p >> to-delete.txt
  fi
done

Delete the artifacts; best run it on ssh jenkins-master for performance.

n=$(cat to-delete.txt | wc -l)
for p in $(cat to-delete.txt); do
  n=$((n-1))
  echo "$p - $n"
  curl -u 'lukas:PASSWORDSEKRET' -X DELETE "https://scala-ci.typesafe.com/artifactory/scala-pr-validation-snapshots/$p"
done

After that

Other measures

</details>

Resize drives / file systems

Enlarging drives and their partitions seems to work well, even for the root partition of a running system (Debian).

Recreate drive

To recreate a drive / file system (to shrink it, or to move to a different file system), create a new EBS volume, mount it and copy the data over using rsync.

Unattended upgrades

Enabled on master and behemoths

JVM installations

Only a basic jre is installed through apt (eg openjdk-17-jre-headless).

Use sudo su and cd to /usr/lib/jvm, see the README file. Install new JDKs here, we default to adoptium.

chrony

On all machines (chronyc tracking to check):

root@ip-172-31-10-237:~# cat /etc/chrony/sources.d/aws.sources
#https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/configure-ec2-ntp.html
server 169.254.169.123 prefer iburst minpoll 4 maxpoll 4