-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Description
Describe the bug
@pmcgleenon reports on #17721 (comment):
The instance type c6a.xlarge only has 8GB RAM (the clickbench dataset is 15GB) and there are a number of issues with it, including
...
OOM errors happens during test execution, with results reported as null for several tests
To Reproduce
- Use docker to build datafusion:
cd .devcontainer
docker build -t datafusion-build .
cd ..
docker run -m 4G -v `pwd`:/datafusion -it datafusion-build /bin/bashNow, in the docker container
# build datafusion-cli
cd /datafusion
cargo install --profile=release-nonlto --path datafusion-cli
# Get benchmark data
cd /datafusion/benchmarks
./bench.sh data clickbench_partitioned
cd /datafusion/benchmarks/data
# make symlink to hits so queries can run without modification
ln -s hits_partitioned hits
# run the queries
for q in `ls ../queries/clickbench/queries/*.sql` ; do datafusion-cli -f $q ; doneThis is the loop that runs the queries
for q in `ls ../queries/clickbench/queries/*.sql` ; do
echo "Running $q..." ;
datafusion-cli -f $q ;
doneYou'll see queries get killed due to OOM like this:
Running ../queries/clickbench/queries/q18.sql...
DataFusion CLI v50.3.0
bash: line 1: 57537 Killed datafusion-cli -f $q
The queries that are killed are:
Running ../queries/clickbench/queries/q18.sql...
Running ../queries/clickbench/queries/q20.sql...
Running ../queries/clickbench/queries/q22.sql...
Running ../queries/clickbench/queries/q23.sql...
Running ../queries/clickbench/queries/q32.sql...
Running ../queries/clickbench/queries/q33.sql...
Running ../queries/clickbench/queries/q34.sql...
Running ../queries/clickbench/queries/q35.sql...
You can find these queries here: https://github.com/apache/datafusion/tree/main/benchmarks/queries/clickbench/queries
Expected behavior
The queries should not be killed
Additional context
This ticket tracks build OOM'ing: