-
Notifications
You must be signed in to change notification settings - Fork 91
/
container-entrypoint.sh
executable file
·77 lines (69 loc) · 2.27 KB
/
container-entrypoint.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
#! /bin/bash
# Copyright (c) 2024, NVIDIA CORPORATION. All rights reserved.
#
# Licensed 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.
# Start the scheduler on the rank 0 node
if [[ -z "$SLURM_NODEID" ]] || [[ $SLURM_NODEID == 0 ]]; then
# Make the directories needed
echo "Making log directory $LOGDIR"
mkdir -p $LOGDIR
echo "Making profile directory $PROFILESDIR"
mkdir -p $PROFILESDIR
echo "Starting scheduler"
if [[ $DEVICE == 'cpu' ]]; then
dask scheduler \
--scheduler-file $SCHEDULER_FILE \
--protocol $PROTOCOL \
--interface $INTERFACE >> $SCHEDULER_LOG 2>&1 &
fi
if [[ $DEVICE == 'gpu' ]]; then
DASK_DISTRIBUTED__COMM__UCX__CREATE_CUDA_CONTEXT=True \
DASK_DISTRIBUTED__RMM__POOL_SIZE=$RMM_SCHEDULER_POOL_SIZE \
dask scheduler \
--scheduler-file $SCHEDULER_FILE \
--protocol $PROTOCOL \
--interface $INTERFACE >> $SCHEDULER_LOG 2>&1 &
fi
fi
# Wait for the scheduler to start
sleep 30
# Start the workers on each node
echo "Starting workers..."
export WORKER_LOG=$LOGDIR/worker_${SLURM_NODEID}-${SLURM_LOCALID}.log
if [[ $DEVICE == 'cpu' ]]; then
dask worker \
--scheduler-file $SCHEDULER_FILE \
--memory-limit $CPU_WORKER_MEMORY_LIMIT \
--nworkers -1 \
--interface $INTERFACE >> $WORKER_LOG 2>&1 &
fi
if [[ $DEVICE == 'gpu' ]]; then
dask-cuda-worker \
--scheduler-file $SCHEDULER_FILE \
--rmm-pool-size $RMM_WORKER_POOL_SIZE \
--interface $INTERFACE \
--enable-cudf-spill \
--rmm-async >> $WORKER_LOG 2>&1 &
fi
# Wait for the workers to start
sleep 60
if [[ -z "$SLURM_NODEID" ]] || [[ $SLURM_NODEID == 0 ]]; then
echo "Starting $SCRIPT_COMMAND"
bash -c "$SCRIPT_COMMAND"
touch $DONE_MARKER
fi
# All nodes wait until done to keep the workers and scheduler active
while [ ! -f $DONE_MARKER ]
do
sleep 15
done