Clarify meaning the difference between `command.run` timeouts and request_timeout

[brendanator]

Describe the bug
I don't understand what commands.run timeouts are supposed to do

To Reproduce
Steps to reproduce the behavior:

Using timeout an exception is raised but the command continues to run:

from time import sleep
from e2b_code_interpreter import Sandbox
import logging

logger = logging.getLogger(__name__)

with Sandbox() as sbx:
    try:
        logger.info("Running sleep 10")
        sbx.commands.run("sleep 10", timeout=1)
        logger.info("Sleep 10 finished")
    except Exception as e:
        logger.error(e)
    for i in range(10):
        logger.info("%d seconds", i)
        logger.info(sbx.commands.list())
        sleep(1)

2024-12-11 09:30:06,703 - INFO - Running sleep 10
2024-12-11 09:31:06,913 - ERROR - context deadline exceeded: This error is likely due to exceeding 'timeoutMs' — the total time a long running request (like process or directory watch) can be active. It can be modified by passing 'timeoutMs' when making the request. Use '0' to disable the timeout.
2024-12-11 09:31:06,913 - INFO - 0 seconds
2024-12-11 09:31:07,077 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:08,082 - INFO - 1 seconds
2024-12-11 09:31:08,202 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:09,206 - INFO - 2 seconds
2024-12-11 09:31:09,321 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:10,327 - INFO - 3 seconds
2024-12-11 09:31:10,450 - INFO - [ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd=''), ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:31:11,451 - INFO - 4 seconds
2024-12-11 09:31:11,574 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:12,579 - INFO - 5 seconds
2024-12-11 09:31:12,693 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:13,698 - INFO - 6 seconds
2024-12-11 09:31:13,818 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:14,823 - INFO - 7 seconds
2024-12-11 09:31:14,940 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user'), ProcessInfo(pid=922, tag='', cmd='/bin/bash', args=['-l', '-c', 'sleep 10'], envs={}, cwd='')]
2024-12-11 09:31:15,945 - INFO - 8 seconds
2024-12-11 09:31:16,063 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:31:17,068 - INFO - 9 seconds
2024-12-11 09:31:17,179 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]

Using request_timeout does not appear to do anything:

from time import sleep
from e2b_code_interpreter import Sandbox
import logging

logger = logging.getLogger(__name__)

with Sandbox() as sbx:
    try:
        logger.info("Running sleep 10")
        sbx.commands.run("sleep 10", request_timeout=1)  # <- changed line
        logger.info("Sleep 10 finished")
    except Exception as e:
        logger.error(e)
    for i in range(10):
        logger.info("%d seconds", i)
        logger.info(sbx.commands.list())
        sleep(1)

2024-12-11 09:34:15,703 - INFO - Running sleep 10
2024-12-11 09:34:25,895 - INFO - Sleep 10 finished
2024-12-11 09:34:25,896 - INFO - 0 seconds
2024-12-11 09:34:26,010 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:27,016 - INFO - 1 seconds
2024-12-11 09:34:27,132 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:28,137 - INFO - 2 seconds
2024-12-11 09:34:28,252 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:29,257 - INFO - 3 seconds
2024-12-11 09:34:29,373 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:30,376 - INFO - 4 seconds
2024-12-11 09:34:30,489 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:31,489 - INFO - 5 seconds
2024-12-11 09:34:31,602 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:32,607 - INFO - 6 seconds
2024-12-11 09:34:32,723 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:33,728 - INFO - 7 seconds
2024-12-11 09:34:33,841 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:34,846 - INFO - 8 seconds
2024-12-11 09:34:34,960 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2024-12-11 09:34:35,965 - INFO - 9 seconds
2024-12-11 09:34:36,084 - INFO - [ProcessInfo(pid=580, tag='startCmd', cmd='/bin/bash', args=['-l', '-c', '/root/.jupyter/start-up.sh'], envs={}, cwd='/home/user')]
2

Expected behavior

• I would expect the job to be killed when timeout is exceeded
• I would expect something to happen when request_timeout is exceeded

Maybe I'm misunderstanding what these are expected to do. Either way I don't understand the docs

timeout: Timeout for the command connection in seconds. Using 0 will not limit the command connection time
request_timeout: Timeout for the request in seconds

Questions:

• What is the expected behaviour of timeout and request_timeout? What exceptions can they throw if they are exceeded? And what happens to the running command when they are exceeded?
• What is the expected behaviour if the Sandbox timeout is exceeded? Is the Sandbox killed? Are running command killed? What happens to subsequent attempts to run command or read files?
• What is the expected behaviour if sbx.files.write request_timeout is exceeded? When can this happen? Will the write attempt continue to run?

Thanks!

[linear]

E2B-1289 Clarify meaning the difference between `command.run` timeouts and request_timeout

[ValentaTomas]

Hey @brendanator, sorry for the late response.
The request_timeout is the timeout for making the request to the sandbox. The actual timeout is how long the long running operation (in this case the running command) can take. They are separated because often you want to throw an error if you cannot start the operation, but then you want to keep the operation running for a long time.

What is the expected behaviour of timeout and request_timeout? What exceptions can they throw if they are exceeded? And what happens to the running command when they are exceeded?

They should be throwing the TimeoutException. Right now describe what happened in the exception message, but we plan to separate these by types.

What is the expected behaviour if the Sandbox timeout is exceeded? Is the Sandbox killed? Are running command killed? What happens to subsequent attempts to run command or read files?

After sandbox times out it is killed and all request to it fail. Right now long running requests like commands might take a little time to be notified about the sandbox end.

What is the expected behaviour if sbx.files.write request_timeout is exceeded? When can this happen? Will the write attempt continue to run?

If you exceed the request timeout of the .write, the request gets cancelled and the write attempt with it.