88 lines
3.2 KiB
Bash
Executable File
88 lines
3.2 KiB
Bash
Executable File
#!/bin/bash
|
|
# Per-GPU work queue worker, fed from the S3 bucket (rclone remote "jspace").
|
|
#
|
|
# Usage: gpuq_worker.sh <worker-id> <gpu-index>
|
|
# e.g.: gpuq_worker.sh node1-gpu3 3
|
|
# Run ONE per GPU, each in its own tmux window:
|
|
# tmux new-window -t ssh_tmux -n gpuq3 'bash ~/jspace/scripts/gpuq_worker.sh node1-gpu3 3'
|
|
#
|
|
# Queue layout (per worker, race-free by construction — no work stealing):
|
|
# jspace:jspace/gpuq/<worker-id>/pending/<job>.sh submitted jobs
|
|
# jspace:jspace/gpuq/<worker-id>/done/<job>.sh|.log finished jobs + logs
|
|
# Jobs run oldest-first by name — prefix with 00_, 01_, ... to order them.
|
|
#
|
|
# Job contract: plain bash, repo at $GPUQ_REPO (default ~/jspace),
|
|
# CUDA_VISIBLE_DEVICES is set by the worker.
|
|
#
|
|
# DATA CONTRACT (selective sync — declare in job-file comments):
|
|
# # gpuq-in: results-loop/mbpp_data.json results-loop/adapter_code.pt
|
|
# # gpuq-out: results-loop/eval_code_gate*.json results/jbar_31b.pt
|
|
# Paths are repo-relative and mirror to jspace:jspace/<same path> in the
|
|
# bucket. Inputs are pulled before the job starts; outputs are pushed
|
|
# every SYNC_EVERY seconds while the job runs AND once at exit — a dying
|
|
# node loses at most one sync interval. Globs allowed in the basename.
|
|
# Repeat the comment lines to declare more paths. No declaration = no sync
|
|
# (old jobs keep working; they can still rclone by hand).
|
|
#
|
|
# GRACEFUL RESTART: touch $HOME/gpuq/<worker-id>/STOP — the worker exits
|
|
# between jobs, never mid-job. NEVER kill the tmux session while a job
|
|
# runs (that killed a 3.5h scan once).
|
|
|
|
set -u
|
|
ID=$1
|
|
GPU=$2
|
|
Q="jspace:jspace/gpuq/$ID"
|
|
LOCAL="$HOME/gpuq/$ID"
|
|
REPO="${GPUQ_REPO:-$HOME/jspace}"
|
|
SYNC_EVERY="${GPUQ_SYNC_EVERY:-120}"
|
|
mkdir -p "$LOCAL/pending"
|
|
|
|
sync_paths() { # $1 = direction (in|out), $2 = job file
|
|
local dir line p
|
|
while IFS= read -r line; do
|
|
for p in $line; do
|
|
local d b
|
|
d=$(dirname "$p"); b=$(basename "$p")
|
|
mkdir -p "$REPO/$d"
|
|
if [ "$1" = in ]; then
|
|
rclone copy "jspace:jspace/$d" "$REPO/$d" --include "$b" -q 2>/dev/null
|
|
else
|
|
rclone copy "$REPO/$d" "jspace:jspace/$d" --include "$b" -q 2>/dev/null
|
|
fi
|
|
done
|
|
done < <(sed -n "s/^# gpuq-$1:[[:space:]]*//p" "$2")
|
|
}
|
|
|
|
echo "[gpuq $ID] worker up on GPU $GPU, polling $Q/pending (repo $REPO)"
|
|
while true; do
|
|
if [ -e "$LOCAL/STOP" ]; then
|
|
rm -f "$LOCAL/STOP"
|
|
echo "[gpuq $ID] $(date +%F_%H:%M:%S) STOP sentinel — draining (no job running)"
|
|
break
|
|
fi
|
|
rclone move "$Q/pending" "$LOCAL/pending" --include "*.sh" -q 2>/dev/null
|
|
job=$(ls "$LOCAL"/pending/*.sh 2>/dev/null | sort | head -1)
|
|
if [ -z "$job" ]; then
|
|
sleep 20
|
|
continue
|
|
fi
|
|
name=$(basename "$job" .sh)
|
|
echo "[gpuq $ID] $(date +%F_%H:%M:%S) START $name"
|
|
sync_paths in "$job"
|
|
CUDA_VISIBLE_DEVICES=$GPU bash "$job" > "$LOCAL/$name.log" 2>&1 &
|
|
pid=$!
|
|
while kill -0 $pid 2>/dev/null; do
|
|
sleep "$SYNC_EVERY"
|
|
sync_paths out "$job"
|
|
rclone copy "$LOCAL/$name.log" "$Q/running/" -q 2>/dev/null
|
|
done
|
|
wait $pid
|
|
rc=$?
|
|
echo "[gpuq $ID] $(date +%F_%H:%M:%S) DONE $name rc=$rc"
|
|
echo "rc=$rc" >> "$LOCAL/$name.log"
|
|
sync_paths out "$job"
|
|
rclone copy "$LOCAL/$name.log" "$Q/done/" -q
|
|
rclone deletefile "$Q/running/$name.log" -q 2>/dev/null
|
|
rclone moveto "$job" "$Q/done/$name.sh" -q
|
|
done
|