cassandra/.circleci/readme.md

196 lines
9.8 KiB
Markdown

<!--
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
-->
# CircleCI config files
This directory contains the configuration for CircleCI continous integration platform.
The file `config.yml` is the configuration file that is read by CircleCI. This file is
automatically generated by the `generate.sh` script from the `config_template.yml` file.
The provided `config.yml` file uses low resources so users of the CircleCI free tier can
use it. Additionally, there are two versions of this file using different resources so
users who have access to premium CircleCI resources can use larger instances and more
parallelism. These files are `config.yml.FREE` and `config.yml.PAID`.
The default `config.yml` file is just a copy of `config.yml.FREE`.
## Switching to higher resource settings
This directory contains generated files for free and paid resource settings.
Switch between them by copying the correct file to `config.yml` and committing the result.
For example, for using higher resources benefiting from CircleCI's paid tier:
`cp .circleci/config.yml.PAID .circleci/config.yml`
And for using lower resources comaptible with CircleCI's free tier:
`cp .circleci/config.yml.FREE .circleci/config.yml`
Alternatively, you can run the `generate.sh` script with the flags `-f`/`-p`
to regenerate the `config.yml` file from `config_template.yml` using free or paid resources.
The script validates and applies any changes to the `config_template.yml` file, and they require the
[CircleCI CLI](https://circleci.com/docs/2.0/local-cli/#install) to be installed.
## Setting environment variables
Both `config_template.yml` and `config.yml` files contain a set of environment variables
defining things like what dtest repo and branch to use, what tests could be repeatedly
run, etc.
These environment variables can be directly edited in the `config.yml` file, although if
you do this you should take into account that the entire set of env vars is repeated on
every job.
A probably better approach is editing them in `config_template.yml` and then regenerate the `config.yml`
files using the `generate.sh` script. You can also
directly pass environment variable values to the `generate.sh` script
with the `-e` flag. For example, to set the dtest repo and branch with PAID config
you can run:
```
generate.sh -p \
-e DTEST_REPO=https://github.com/adelapena/cassandra-dtest.git \
-e DTEST_BRANCH=CASSANDRA-8272
```
## Running tests in a loop
Running the `generate.sh` script will use `git diff` to find the new or modified tests.
The scripts will then create jobs to run each of these new or modified tests for a certain number
of times, to verify that they are stable. You can use environment variables to specify the number of
iterations of each type of test:
```
generate.sh -p \
-e REPEATED_UTESTS_COUNT=500 \
-e REPEATED_UTESTS_FQLTOOL_COUNT=500 \
-e REPEATED_UTESTS_LONG_COUNT=100 \
-e REPEATED_UTESTS_STRESS_COUNT=500 \
-e REPEATED_SIMULATOR_DTESTS_COUNT=500 \
-e REPEATED_JVM_DTESTS_COUNT=500 \
-e REPEATED_JVM_UPGRADE_DTESTS_COUNT=500 \
-e REPEATED_DTESTS_COUNT=500 \
-e REPEATED_LARGE_DTESTS_COUNT=100 \
-e REPEATED_UPGRADE_DTESTS_COUNT=25 \
-e REPEATED_ANT_TEST_COUNT=500
```
You can also specify whether the iteration should fail on the first test failure:
```
generate.sh -p -e REPEATED_TESTS_STOP_ON_FAILURE=false
```
In addition to the automatically detected tests, it's also possible to provide lists of
specific tests to be repeated:
```
generate.sh -p \
-e REPEATED_UTESTS=org.apache.cassandra.cql3.ViewTest,org.apache.cassandra.db.CellTest \
-e REPEATED_UTESTS_FQLTOOL=org.apache.cassandra.fqltool.FQLCompareTest \
-e REPEATED_UTESTS_LONG=org.apache.cassandra.io.sstable.CQLSSTableWriterLongTest#testWideRow \
-e REPEATED_UTESTS_STRESS=org.apache.cassandra.stress.generate.DistributionGaussianTest \
-e REPEATED_SIMULATOR_DTESTS=org.apache.cassandra.simulator.test.TrivialSimulationTest \
-e REPEATED_DTESTS=cql_test.py,consistency_test.py::TestAvailability::test_simple_strategy \
-e REPEATED_LARGE_DTESTS=replace_address_test.py::TestReplaceAddress::test_replace_stopped_node \
-e REPEATED_JVM_DTESTS=org.apache.cassandra.distributed.test.PagingTest#testPaging \
-e REPEATED_UPGRADE_DTESTS=upgrade_tests/cql_tests.py \
-e REPEATED_JVM_UPGRADE_DTESTS=org.apache.cassandra.distributed.upgrade.GroupByTest
```
If you need to repeat a particular set of tests and avoid automatic detection of changed tests
use the -s command line option.
For particular Ant test targets that are not included in the regular test suites, you can
use the `run_repeated_utest` job:
```
generate.sh -p \
-e REPEATED_ANT_TEST_TARGET=test-cdc \
-e REPEATED_ANT_TEST_CLASS=org.apache.cassandra.cql3.ViewTest \
-e REPEATED_ANT_TEST_METHODS=testCompoundPartitionKey,testStaticTable \
-e REPEATED_ANT_TEST_VNODES=false \
-e REPEATED_ANT_TEST_COUNT=500
```
Putting all together, you can have runs as complex as:
```
generate.sh -p \
-e REPEATED_TESTS_STOP_ON_FAILURE=true \
-e REPEATED_UTESTS=org.apache.cassandra.cql3.ViewTest,org.apache.cassandra.db.CellTest \
-e REPEATED_UTESTS_COUNT=500 \
-e REPEATED_UTESTS_FQLTOOL=org.apache.cassandra.fqltool.FQLCompareTest \
-e REPEATED_UTESTS_FQLTOOL_COUNT=500 \
-e REPEATED_UTESTS_LONG=org.apache.cassandra.io.sstable.CQLSSTableWriterLongTest#testWideRow \
-e REPEATED_UTESTS_LONG_COUNT=100 \
-e REPEATED_UTESTS_STRESS=org.apache.cassandra.stress.generate.DistributionGaussianTest \
-e REPEATED_UTESTS_STRESS_COUNT=500 \
-e REPEATED_SIMULATOR_DTESTS=org.apache.cassandra.simulator.test.TrivialSimulationTest \
-e REPEATED_SIMULATOR_DTESTS_COUNT=500 \
-e REPEATED_DTESTS=cql_test.py,consistency_test.py::TestAvailability::test_simple_strategy \
-e REPEATED_DTESTS_COUNT=500 \
-e REPEATED_LARGE_DTESTS=replace_address_test.py,materialized_views_test.py \
-e REPEATED_LARGE_DTESTS_COUNT=100 \
-e REPEATED_JVM_DTESTS=org.apache.cassandra.distributed.test.PagingTest#testPaging \
-e REPEATED_JVM_DTESTS_COUNT=500 \
-e REPEATED_UPGRADE_DTESTS=upgrade_tests/cql_tests.py \
-e REPEATED_UPGRADE_DTESTS_COUNT=25 \
-e REPEATED_JVM_UPGRADE_DTESTS=org.apache.cassandra.distributed.upgrade.GroupByTest \
-e REPEATED_JVM_UPGRADE_DTESTS_COUNT=500 \
-e REPEATED_ANT_TEST_TARGET=test-cdc \
-e REPEATED_ANT_TEST_CLASS=org.apache.cassandra.cql3.ViewTest \
-e REPEATED_ANT_TEST_METHODS=testCompoundPartitionKey,testStaticTable \
-e REPEATED_ANT_TEST_VNODES=false \
-e REPEATED_ANT_TEST_COUNT=500
```
WARNING: The Python large DTests test_network_topology_strategy and test_network_topology_strategy_each_quorum require
XLarge containers in order to be run in a loop. Otherwise, you would see NO HOST AVAILABLE if you run them with the
default medium config and Large containers.
## Updating the config
For configuration changes meant to be permanent in the Apache repo you should never edit
the `config.yml` file manually. Instead, you should edit the `config_template.yml` file and then
regenerate the `config.yml`, `config.yml.FREE` and `config.yml.PAID`
files by runnining the `generate.sh` script with `-a` flag. For using this script you
need to install the [CircleCI CLI](https://circleci.com/docs/2.0/local-cli/#install).
As for temporal changes done while working in a patch, such as pointing to you dtest repo or
running a test repeatedly, you can either directly edit `config.yml` or edit `config_template.yml`
and then regenerate `config.yml` with the `generate.sh` script using a `-f`/`-p` flag.
When this flag is used only the `config.yml` will be generated. Same workflow applies to the respective
JDK11+17 CircleCI configuration files.
Please note that any previous swapping or edition of the generated files will be overriden
by running `generate.sh` with `-a` argument, returning `config.yml`to the default FREE.
So if you previously swapped your `config.yml` to PAID you would need to either swap it again
or use the `-f`/`-p` script flags.
Read below for details how to generate the files manually without the `generate.sh` script:
1. make your edits to config_template.yml - let it stay at free tier settings
2. generate a valid FREE file:
`circleci config process config_template.yml > config.yml.FREE`
3. add the Apache license header to the newly created FREE file:
`cat license.yml config.yml.FREE > config.yml.FREE.new && mv config.yml.FREE.new config.yml.FREE`
4. then apply the PAID patch to config_template.yml;
`patch -o config_template.yml.PAID config_template.yml config_template.yml.PAID.patch`
(this creates a new file `config_template.yml.PAID` instead of in-place patching
config_template.yml).
Note that if the patch no longer applies to `config_template.yml` a new patch file
is needed, do this by manually creating and editing `config_template.yml.PAID`, and create
the patch file based on the diff (don't commit it though).
5. generate the PAID file:
`circleci config process config_template.yml.PAID > config.yml.PAID`
6. remove the temporary patched PAID file: `rm config_template.yml.PAID`
7. add the Apache license header to the newly created PAID file:
`cat license.yml config.yml.PAID > config.yml.PAID.new && mv config.yml.PAID.new config.yml.PAID`
8. finally, remember to update the config.yml