Skip to content

try partitions #6834

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
SteffenDE wants to merge 2 commits into
base: main
Choose a base branch
from
+36 −5
Open

try partitions #6834

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
97c2378
try partitions
SteffenDE Sep 4, 2026
caa5a5d
give db init more time
SteffenDE Sep 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 28 additions & 4 deletions .github/workflows/ci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -223,13 +223,20 @@ jobs:
run: mix docs --warnings-as-errors

integration-test-elixir:
name: integration test (OTP ${{matrix.otp}} | Elixir ${{matrix.elixir}})
name: integration test (OTP ${{matrix.otp}} | Elixir ${{matrix.elixir}} | partition ${{matrix.partition}})

runs-on: ubuntu-24.04
timeout-minutes: 30

strategy:
# Keep running the remaining partitions when one fails, otherwise a single
# failing partition hides the results of all the others.

@rhcarvalho rhcarvalho Sep 4, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This was something I thought about this week. The previous approach was abort the other Elixir/OTP if one fails.

With or without partitions, fail-fast: false gives us complete visibility trading off CPU time. The only constraint here is 2,000 free Actions minutes per month:

fail-fast: false
matrix:
# Test files are split across Elixir instances via `mix test --partitions`.
# Keep this list in sync with the `--partitions` count in the test step below.
partition: [1, 2, 3]
elixir: ["1.18.4", "1.20.4"]
include:
- elixir: "1.18.4"
otp: "27.3.4.3"
Comment on lines +239 to 242

@rhcarvalho rhcarvalho Sep 4, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you do something like this we only need to write the versions once (I have something like this in a local branch):

Suggested change
elixir: ["1.18.4", "1.20.4"]
include:
- elixir: "1.18.4"
otp: "27.3.4.3"
versions:
- elixir: "1.18.4"
otp: "27.3.4.3"

Then update the usage sites to be matrix.versions.elixir, etc.

Makes it easier to bump the versions without accidents.

Expand All @@ -240,6 +247,7 @@ jobs:
env:
ELIXIR_ASSERT_TIMEOUT: 10000
PHX_CI: true
MIX_TEST_PARTITION: ${{ matrix.partition }}
# Set to half of the runner vCPU cores to compile deps/NIFs
# concurrently across OS processes without CPU oversubscription.
# See https://docs.github.com/en/actions/reference/runners/github-hosted-runners#standard-github-hosted-runners-for-public-repositories
Expand All @@ -254,22 +262,38 @@ jobs:
- 5432:5432
env:
POSTGRES_PASSWORD: postgres
# Probe over TCP (-h 127.0.0.1) rather than the Unix socket: initdb runs
# a temporary server with `listen_addresses` empty, and a socket probe
# reports ready against *that* server, before the real one is listening.
options: >-
--health-cmd "pg_isready -U postgres"
--health-cmd "pg_isready -U postgres -h 127.0.0.1"
--health-interval 2s
--health-timeout 3s
--health-retries 10
--health-start-period 30s
mysql:
image: mysql:26
ports:
- 3306:3306
env:
MYSQL_ALLOW_EMPTY_PASSWORD: "yes"
# Things to be aware of:
#
# 1. Initializing the data directory on first boot takes 30s+ on a slow
# runner, which overruns the retry budget below. Probes that fail
# during --health-start-period do not consume that budget, and the
# first successful probe ends the period early.
# 2. The entrypoint starts a temporary server during initialization with
# `port: 0` (Unix socket only). `mysqladmin ping -h localhost` uses
# the socket, so it reports healthy against that temporary server --
# which is then shut down and restarted. Probing over TCP instead
# only succeeds once the real server is listening on 3306.
options: >-
--health-cmd "mysqladmin ping -h localhost"
--health-cmd "mysqladmin ping -h 127.0.0.1 --protocol=TCP"
--health-interval 2s
--health-timeout 3s
--health-retries 10
--health-start-period 60s
mssql:
image: mcr.microsoft.com/mssql/server:2019-latest
env:
Expand Down Expand Up @@ -323,6 +347,6 @@ jobs:
- name: Run integration tests
working-directory: integration_test
run: |
mix test --include database \
mix test --include database --partitions 3 \
--formatter ExUnit.CLIFormatter \
--formatter Phoenix.Integration.SummaryFormatter
9 changes: 8 additions & 1 deletion integration_test/test/support/summary_formatter.ex
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,7 +130,14 @@ defmodule Phoenix.Integration.SummaryFormatter do
end)
|> Enum.sort_by(& &1.total_us, :desc)

env_info = "Elixir #{System.version()} / OTP #{System.otp_release()}"
partition_info =
case System.get_env("MIX_TEST_PARTITION") do
nil -> ""
"" -> ""
partition -> " / partition #{partition}"
end

env_info = "Elixir #{System.version()} / OTP #{System.otp_release()}#{partition_info}"

sections = [
"""
Expand Down
Loading