I have a job in Kubernetes with a python script that is printing a tqdm progress bar to stdout (or stderr?). However, the progress bar cannot be seen with:
kubectl logs test-qpgb7 # returns nothing
kubectl logs -f test-qpgb7 # hangs and returns nothing also
apiVersion: batch/v1
kind: Job
metadata:
name: test
labels:
job: test
spec:
template:
spec:
containers:
- name: test
image: my-image
imagePullPolicy: IfNotPresent
command: ["python", "test.py"]
restartPolicy: Never
backoffLimit: 1
# test.py
import tqdm
from time import sleep
if __name__ == "__main__":
for i in tqdm.tqdm(range(0, 1000)):
sleep(2)
Is it possible to see the progress bar status while the job has not yet completed? When the job completes
kubectl logs test-qpgb7 does work.
UPDATE: The problem does not seem to be in tqdm since this does not show output either:
# test.py
# import tqdm
from time import sleep
if __name__ == "__main__":
for i in range(0, 1000):
print(str(i))
sleep(2)
FROM python:3.9.1
WORKDIR /integrity
COPY requirements.txt .
COPY src/ .
RUN pip install -r requirements.txt
UPDATE2:
adding a sys.stdout.flush() in the print for loop:
if __name__ == "__main__":
for i in range(0, 1000):
print(str(i))
sys.stdout.flush()
sleep(2)
Does work with just the print but not with the tqdm progress. Note sys.stderr.flush() was also added with no success.
UPDATE3: For now a temporary solution is:
from tqdm import tqdm
from time import sleep
import sys
if __name__ == "__main__":
for i in tqdm(range(0, 1000), file=sys.stdout):
print("")
sys.stdout.flush()
sys.stderr.flush()
sleep(2)
# kubectl logs -f test007-ctqjh
0%| | 0/1000 [00:00<?, ?it/s]
0%| | 1/1000 [00:02<33:20, 2.00s/it]
0%| | 2/1000 [00:04<33:18, 2.00s/it]
0%| | 3/1000 [00:06<33:16, 2.00s/it]
0%| | 4/1000 [00:08<33:14, 2.00s/it]
0%| | 5/1000 [00:10<33:12, 2.00s/it]
1%| | 6/1000 [00:12<33:10, 2.00s/it]
which prints on a new line every time. However, will be interested if someone finds how to make this work without printing to a new line.
