The docker module used different code for socket-activated docker daemon than for the non-socket activated daemon.
In particular, if the socket-activated daemon is used, then modprobe wasn't set up to be usable and in PATH for
the docker daemon, which resulted in a failure to start the daemon with overlayfs as storageDriver if the
`overlay` kernel module wasn't already loaded. This commit fixes that bug (which only appears if socket
activation is used), and also reduces the duplication between code paths so that it's easier to keep
both in sync in future.
As @domenkozar noted in #10828, cache=writeback seems to do more harm
than good:
https://github.com/NixOS/nixpkgs/issues/10828#issuecomment-164426821
He has tested it using the openstack NixOS tests and found that
cache=none significantly improves startup performance.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
This seems to be the root cause of the random page allocation failures
and @wizeman did a very good job on not only finding the root problem
but also giving a detailed explanation of it in #10828.
Here is an excerpt:
The problem here is that the kernel is trying to allocate a contiguous
section of 2^7=128 pages, which is 512 KB. This is way too much:
kernel pages tend to get fragmented over time and kernel developers
often go to great lengths to try allocating at most only 1 contiguous
page at a time whenever they can.
From the error message, it looks like the culprit is unionfs, but this
is misleading: unionfs is the name of the userspace process that was
running when the system ran out of memory, but it wasn't unionfs who
was allocating the memory: it was the kernel; specifically it was the
v9fs_dir_readdir_dotl() function, which is the code for handling the
readdir() function in the 9p filesystem (the filesystem that is used
to share a directory structure between a qemu host and its VM).
If you look at the code, here's what it's doing at the moment it tries
to allocate memory:
buflen = fid->clnt->msize - P9_IOHDRSZ;
rdir = v9fs_alloc_rdir_buf(file, buflen);
If you look into v9fs_alloc_rdir_buf(), you will see that it will try
to allocate a contiguous buffer of memory (using kzalloc(), which is a
wrapper around kmalloc()) of size buflen + 8 bytes or so.
So in reality, this code actually allocates a buffer of size
proportional to fid->clnt->msize. What is this msize? If you follow
the definition of the structures, you will see that it's the
negotiated buffer transfer size between 9p client and 9p server. On
the client side, it can be controlled with the msize mount option.
What this all means is that, the reason for running out of memory is
that the code (which we can't easily change) tries to allocate a
contiguous buffer of size more or less equal to "negotiated 9p
protocol buffer size", which seems to be way too big (in our NixOS
tests, at least).
After that initial finding, @lethalman tested the gnome3 gdm test
without setting the msize parameter at all and it seems to have resolved
the problem.
The reason why I'm committing this without testing against all of the
NixOS VM test is basically that I think we can only go better but not
worse than the current state.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
The most complex problems were from dealing with switches reverted in
the meantime (gcc5, gmp6, ncurses6).
It's likely that darwin is (still) broken nontrivially.
Commit 9bfe92ecee ("docker: Minor improvements, fix failing test") added
the services.docker.storageDriver option, made it mandatory but didn't
give it a default value. This results in an ugly traceback when users
enable docker, if they don't pay enough attention to also set the
storageDriver option. (An attempt was made to add an assertion, but it
didn't work, possibly because of how "mkMerge" works.)
The arguments against a default value were that the optimal value
depends on the filesystem on the host. This is, AFAICT, only in part
true. (It seems some backends are filesystem agnostic.) Also, docker
itself uses a default storage driver, "devicemapper", when no
--storage-driver=x options are given. Hence, we use the same value as
default.
Add a FIXME comment that 'devicemapper' breaks NixOS VM tests (for yet
unknown reasons), so we still run those with the 'overlay' driver.
Closes #10100 and #10217.
When using the ZFS storagedriver in docker, it shells out for the ZFS
commands. The path configuration for the systemd task does not include
ZFS, so if the driver is set to ZFS, add ZFS utilities to the PATH.
This will resolve https://github.com/NixOS/nixpkgs/issues/10127
[Bjørn: prefix commit message with "nixos/docker:", remove extra space
before ';']
The EBS and S3 (instance-store) AMIs are now created from the same
image. HVM instance-store AMIs are also generated.
Disk image generation has been factored out into a function
(nixos/lib/make-disk-image.nix) that can be used to build other kinds
of images.
Booting the demo/installer image won't work if the video memory is too
low. It boots into KDE, shows the background image and doesn't do
anything, according to @domenkozar.
Thanks to @domenkozar for reporting and testing this with 32MB.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
pkgs/os-specific/linux/kernel/common-config.nix defines HIGHMEM64G on
line 441 for 32bit systems, which implies PAE.
We now creating the OVA with PAE support enabled, which fixes bootup of
the image if people are just importing it without setting PAE
explicitly.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
