Tag: tmpfs

multiple random crashes of firefox in a docker container under Linux

Multiple random crashes in Firefox under Linux, when started in a docker container with errors of the kind:

[myuser@92ee57f7f63a ~]$ Exiting due to channel error.
Exiting due to channel error.
[myuser@92ee57f7f63a ~]$ firefox 
ExceptionHandler::GenerateDump cloned child 4864
ExceptionHandler::SendContinueSignalToChild sent continue signal to child
ExceptionHandler::WaitForContinueSignal waiting for continue signal...
Exiting due to channel error.
Exiting due to channel error.
Exiting due to channel error.
Exiting due to channel error.
Bus error (core dumped)
[myuser@92ee57f7f63a ~]$ 
###!!! [Parent][MessageChannel] Error: (msgtype=0x5A001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv
###!!! [Parent][MessageChannel] Error: (msgtype=0x5A001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv
###!!! [Parent][MessageChannel] Error: (msgtype=0x5A001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv
###!!! [Parent][MessageChannel] Error: (msgtype=0x5A001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv
###!!! [Parent][MessageChannel] Error: (msgtype=0x5A001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv

The chances to stop these multiple random crashes if you just increase the memory for the /dev/shm device in the docker container is really good! The default values in the docker /dev/shm are really low (in the tested machine only 64Mbytes) and the recommended values are at least 2Gbytes. More shared tabs more shared memory is needed.

Such strange and random crashes might cause a testing case to fail in an automation testing suite using the Selenium driver to start a Firefox headless instance.

To increase the shared memory size in the docker container, the container should be started with “–shm-size” option. For example:

docker run -it --shm-size=2048m my-gui-fedora:31

If the option –shm-size is missing (probably an old docker software) mapping an already mounted tmpfs directory in the host machine to the container is also a solution:

mkdir /dev/dockershm
mount -t tmpfs -o size=2048m tmpfs /dev/shmdocker
chmod 1777 /dev/shmdocker
docker run -d -v /mnt/shmdocker:/dev/shm my-gui-fedora:31

Remounting in the docker container is also possible:

srv myuser # docker exec -it my-gui bash
[myuser@92ee57f7f63a ~]$ sudo mount -o remount,size=2048m /dev/shm

Note the container should be run with –privileged option to be able to remount the /dev/shm.

The build docker command also has the –shm-size for the purpose of the building process not changing the default shared memory size in the containers based on the image afterward.

tmpfs mount on /dev/shm in LXC container or chroot environment

When using LXC containers booting the lxc container would not populate it as the normal boot process. Or when you create a chroot jail /dev is not mounted or just some devices are created.
There is an option to populate (when using LXC containers) it with minimal required devices:

lxc.autodev = 1

which will create a tmpfs mount under /dev and create some basic devices, it will ensure /dev/shm to be mounted on with tmpfs!
If you omit this option, the /dev directory won’t be populated and will stay with the devices you made or copied when you made the LXC container (or the chroot jail) and /dev/shm will not be mounted using tmps, which could create numerous bad issues.
If you get errors like

 * configure has detected that the sem_open function is broken.
 * Please ensure that /dev/shm is mounted as a tmpfs with mode 1777.

You could mount the /dev/shm of the LXC container or the chroot jail (usually you can tune the size half of the server’s RAM) with

mkdir -p /dev/shm
mount -t tmpfs -o nodev,nosuid,noexec,mode=1777,size=6144m tmpfs /dev/shm

Or reboot your LXC container with a new configuration (probably in the “/var/lxc/[lxc_name]/config”) adding the following line:

lxc.mount.entry = none dev/shm tmpfs nodev,nosuid,noexec,mode=1777,create=dir 0 0

Thus you ensure the /dev/shm to be mounted on tmpfs and all semaphore functions to work properly.

* Real output of Gentoo failed compilation of python package:
 * configure has detected that the sem_open function is broken.
 * Please ensure that /dev/shm is mounted as a tmpfs with mode 1777.
 * ERROR: dev-lang/python-3.3.4-r1::gentoo failed (configure phase):
 *   Broken sem_open function (bug 496328)
 * 
 * Call stack:
 *     ebuild.sh, line 124:  Called src_configure
 *   environment, line 3542:  Called die
 * The specific snippet of code:
 *           die "Broken sem_open function (bug 496328)";
 * 
 * If you need support, post the output of `emerge --info '=dev-lang/python-3.3.4-r1::gentoo'`,
 * the complete build log and the output of `emerge -pqv '=dev-lang/python-3.3.4-r1::gentoo'`.
 * The complete build log is located at '/var/tmp/portage/dev-lang/python-3.3.4-r1/temp/build.log'.
 * The ebuild environment file is located at '/var/tmp/portage/dev-lang/python-3.3.4-r1/temp/environment'.
 * Working directory: '/var/tmp/portage/dev-lang/python-3.3.4-r1/work/x86_64-pc-linux-gnu'
 * S: '/var/tmp/portage/dev-lang/python-3.3.4-r1/work/Python-3.3.4'

>>> Failed to emerge dev-lang/python-3.3.4-r1, Log file: