1.4. Troubleshooting / FAQ
This is a collection of common troubleshooting tips and answers to frequently asked questions.
1.4.1. Getting Help
We love to hear from you. If you have questions, want to discuss an idea, or encountered issues while using SimBricks, we are available on Slack for quick answers and interactive discussions. If you find bugs or want to request a feature, feel free to open an Issue on GitHub.
1.4.2. Error Opening images/output-base/base.raw
Some of our host simulators, e.g., gem5 and Simics, require raw disk images. If these aren’t available, you will see the error in the title or something similar. If you use our Docker images, we deliberately remove these since Docker doesn’t handle large, sparse files well, which leads to large Docker image sizes. We include disk images in the qcow format though, which can easily be converted to raw. To do so, just run the following (requires QEMU to be built first):
$ make convert-images-raw
If you are not using the provided docker containers, you might need to build the qcow images by running the following (again, requires QEMU to be built first):
$ make build-images-min
1.4.3. gem5 Panic PerfKvmCounter::attach
gem5 needs access to the syscall perf_event_open
. To allow access,
/proc/sys/kernel/perf_event_paranoid
has to be set to 1 or lower on the host
system. You can check the current value with
$ cat /proc/sys/kernel/perf_event_paranoid
To change the setting, use this command:
$ sudo sysctl -w kernel.perf_event_paranoid=1
If you are using SimBricks from within a Docker container, you also need to
start the container with the parameter --privileged
for the
perf_event_open
syscall to be accessible. Our VS Code dev container is
configured to already do so, just start it as usual from within VS Code and you
are good to go. Otherwise, add the parameter to your docker run
command.
1.4.4. Is My Simulation Stuck or Just Slow?
It is possible to check the current timestamp of individual component simulators. If the timestamp of a simulator which is synchronizing with at least one other simulator isn’t advancing, the whole simulation is stuck. Many of our component simulators print their timestamp when you send them a USR1 signal, for example, by running
$ kill -s USR1 <insert_pid_of_simulator>
By doing this multiple times, you can check whether the timestamp advances. If
you invoked the
orchestration framework
with --verbose
, the current timestamp is printed directly in the terminal.
If not then you have to stop the experiment via Ctrl+C to produce the output
JSON file. All the simulators’ output is logged there.