Since Jan 2 we've seen a lot of instances of segfaults during conda commands on ursa-i9-9960x. See the short build times below.
Example. These all end with buildkite/benchmark/utils.sh: line 84: <pid> Segmentation fault (core dumped) "$CONDA_EXE" $_CE_M $_CE_CONDA "$cmd" "$@".
I think buildkite/benchmark/utils.sh: line 84 is a red herring because walking through the logs it usually fails during a Solving environment step or around there. In the example I linked I'm fairly certain it's this line:
Since Jan 2 we've seen a lot of instances of segfaults during
conda
commands onursa-i9-9960x
. See the short build times below.Example. These all end with
buildkite/benchmark/utils.sh: line 84: <pid> Segmentation fault (core dumped) "$CONDA_EXE" $_CE_M $_CE_CONDA "$cmd" "$@"
.I think
buildkite/benchmark/utils.sh: line 84
is a red herring because walking through the logs it usually fails during aSolving environment
step or around there. In the example I linked I'm fairly certain it's this line:https://github.com/voltrondata-labs/arrow-benchmarks-ci/blob/87e31d6cae2b9e79ba84cead6d2f259a9832121b/buildkite/benchmark/utils.sh#L10-L17
This isn't happening on the other machines. Wonder if #103 would help.