Skip to content

detect errors in datafusion and fix datafusion vortex on c8g.metal-48xl #1463

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
toschmidt wants to merge 2 commits into
base: main
Choose a base branch
from
Open

detect errors in datafusion and fix datafusion vortex on c8g.metal-48xl #1463

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
0e08032
datafusion(-vortex): fail the query on any stderr output
toschmidt Aug 18, 2026
0a2766a
datafusion-vortex: raise open-file limit so wide scans complete
toschmidt Aug 18, 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
13 changes: 9 additions & 4 deletions datafusion-partitioned/query
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,14 +9,19 @@ DF=arrow-datafusion/target/release/datafusion-cli

query=$(cat)
tmp=$(mktemp /tmp/datafusion.XXXXXX.sql)
trap 'rm -f "$tmp"' EXIT
err=$(mktemp /tmp/datafusion.XXXXXX.err)
trap 'rm -f "$tmp" "$err"' EXIT
printf '%s\n' "$query" > "$tmp"

out=$("$DF" -f create.sql "$tmp" 2>&1) && status=0 || status=$?
# datafusion-cli exits 0 even when a query errors: it prints to stderr and leaves
# only create.sql's "Elapsed" on stdout, which the driver would mis-record as the
# query time. Treat any stderr output (or a nonzero exit) as failure.
out=$("$DF" -f create.sql "$tmp" 2>"$err") && status=0 || status=$?

if [ "$status" -ne 0 ]; then
if [ "$status" -ne 0 ] || [ -s "$err" ]; then
cat "$err" >&2
printf '%s\n' "$out" >&2
exit "$status"
exit 1
fi

printf '%s\n' "$out" | grep -v 'Elapsed' || true
Expand Down
17 changes: 13 additions & 4 deletions datafusion-vortex-partitioned/query
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,16 +5,25 @@ set -e

DF=vortex-datafusion-cli/target/release/vortex-datafusion-cli

# vortex-io opens all partitions at once; on many-core hosts a wide scan can
# exceed the default 1024 open-file limit. Raise the soft limit to the hard cap.
ulimit -n "$(ulimit -Hn)" 2>/dev/null || true

query=$(cat)
tmp=$(mktemp /tmp/datafusion-vortex.XXXXXX.sql)
trap 'rm -f "$tmp"' EXIT
err=$(mktemp /tmp/datafusion-vortex.XXXXXX.err)
trap 'rm -f "$tmp" "$err"' EXIT
printf '%s\n' "$query" > "$tmp"

out=$("$DF" -f create.sql -f "$tmp" 2>&1) && status=0 || status=$?
# The CLI exits 0 even when a query errors: it prints to stderr and leaves only
# create.sql's tiny "Elapsed" on stdout, which the driver would mis-record as the
# query time. Treat any stderr output (or a nonzero exit) as failure.
out=$("$DF" -f create.sql -f "$tmp" 2>"$err") && status=0 || status=$?

if [ "$status" -ne 0 ]; then
if [ "$status" -ne 0 ] || [ -s "$err" ]; then
cat "$err" >&2
printf '%s\n' "$out" >&2
exit "$status"
exit 1
fi

printf '%s\n' "$out" | grep -v 'Elapsed' || true
Expand Down
236 changes: 236 additions & 0 deletions datafusion-vortex-partitioned/results/20260818/c8g.metal-48xl.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,236 @@
{
"system": "DataFusion (Vortex, partitioned)",
"date": "2026-08-18",
"machine": "c8g.metal-48xl",
"cluster_size": 1,
"proprietary": "no",
"hardware": "cpu",
"tuned": "no",
"tags": [
"Rust",
"column-oriented",
"embedded",
"stateless"
],
"load_time": 62,
"data_size": 15329147920,
"concurrent_qps": null,
"concurrent_error_ratio": null,
"result": [
[
0.036,
0.002,
0.002
],
[
0.1,
0.041,
0.042
],
[
0.117,
0.055,
0.059
],
[
0.479,
0.06,
0.065
],
[
0.962,
0.175,
0.158
],
[
1.012,
0.15,
0.15
],
[
0.033,
0.002,
0.002
],
[
0.104,
0.052,
0.052
],
[
0.856,
0.191,
0.191
],
[
1.428,
0.261,
0.242
],
[
0.572,
0.089,
0.093
],
[
1.008,
0.109,
0.098
],
[
1.153,
0.137,
0.134
],
[
2.543,
0.231,
0.242
],
[
0.939,
0.136,
0.141
],
[
0.768,
0.152,
0.143
],
[
2.475,
0.252,
0.262
],
[
2.227,
0.265,
0.279
],
[
3.413,
0.507,
0.522
],
[
0.243,
0.064,
0.046
],
[
14.054,
0.185,
0.193
],
[
16.44,
0.186,
0.197
],
[
22.192,
0.546,
0.224
],
[
53.545,
1.727,
1.403
],
[
1.449,
0.061,
0.054
],
[
1.266,
0.08,
0.082
],
[
1.926,
0.062,
0.071
],
[
14.394,
0.189,
0.202
],
[
11.641,
1.596,
1.926
],
[
0.219,
0.116,
0.124
],
[
2.267,
0.141,
0.14
],
[
5.469,
0.191,
0.193
],
[
3.792,
0.742,
0.747
],
[
14.043,
0.467,
0.466
],
[
14.038,
0.444,
0.457
],
[
0.395,
0.191,
0.197
],
[
0.161,
0.066,
0.067
],
[
0.135,
0.043,
0.044
],
[
0.187,
0.031,
0.03
],
[
0.31,
0.126,
0.124
],
[
0.127,
0.029,
0.03
],
[
0.122,
0.03,
0.03
],
[
0.112,
0.028,
0.029
]
]
}
17 changes: 13 additions & 4 deletions datafusion-vortex/query
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,16 +5,25 @@ set -e

DF=vortex-datafusion-cli/target/release/vortex-datafusion-cli

# vortex-io opens files concurrently; on many-core hosts a wide scan can exceed
# the default 1024 open-file limit. Raise the soft limit to the hard cap.
ulimit -n "$(ulimit -Hn)" 2>/dev/null || true

query=$(cat)
tmp=$(mktemp /tmp/datafusion-vortex.XXXXXX.sql)
trap 'rm -f "$tmp"' EXIT
err=$(mktemp /tmp/datafusion-vortex.XXXXXX.err)
trap 'rm -f "$tmp" "$err"' EXIT
printf '%s\n' "$query" > "$tmp"

out=$("$DF" -f create.sql -f "$tmp" 2>&1) && status=0 || status=$?
# The CLI exits 0 even when a query errors: it prints to stderr and leaves only
# create.sql's tiny "Elapsed" on stdout, which the driver would mis-record as the
# query time. Treat any stderr output (or a nonzero exit) as failure.
out=$("$DF" -f create.sql -f "$tmp" 2>"$err") && status=0 || status=$?

if [ "$status" -ne 0 ]; then
if [ "$status" -ne 0 ] || [ -s "$err" ]; then
cat "$err" >&2
printf '%s\n' "$out" >&2
exit "$status"
exit 1
fi

printf '%s\n' "$out" | grep -v 'Elapsed' || true
Expand Down
Loading