Getting substitute information using the binary cache substituter has
non-trivial latency overhead. A package or NixOS system configuration
can have hundreds of dependencies, and in the worst case (when the
local info cache is empty) we have to do a separate HTTP request for
each of these. If the ping time to the server is t, getting N info
files will take tN seconds; e.g., with a ping time of 0.1s to
nixos.org, sequentially downloading 1000 info files (a typical NixOS
config) will take at least 100 seconds.
To fix this problem, the binary cache substituter can now perform
requests in parallel. This required changing the substituter
interface to support a function querySubstitutablePathInfos() that
queries multiple paths at the same time, and rewriting queryMissing()
to take advantage of parallelism. (Due to local caching,
parallelising queryMissing() is sufficient for most use cases, since
it's almost always called before building a derivation and thus fills
the local info cache.)
For example, parallelism speeds up querying all 1056 paths in a
particular NixOS system configuration from 116s to 2.6s. It works so
well because the eccentricity of the top-level derivation in the
dependency graph is only 9. So we only need 10 round-trips (when
using an unlimited number of parallel connections) to get everything.
Currently we do a maximum of 150 parallel connections to the server.
Thus it's important that the binary cache server (e.g. nixos.org) has
a high connection limit. Alternatively we could use HTTP pipelining,
but WWW::Curl doesn't support it and libcurl has a hard-coded limit of
5 requests per pipeline.
Using WWW::Curl rather than running an external curl process for every
NAR info file halves the time it takes to get info thanks to libcurl's
support for persistent HTTP connections. (We save a roundtrip per
file.) But the real gain will come from using parallel and/or
pipelined requests.
XZ compresses significantly better than bzip2. Here are the
compression ratios and execution times (using 4 cores in parallel) on
my /var/run/current-system (3.1 GiB):
bzip2: total compressed size 849.56 MiB, 30.8% [2m08]
xz -6: total compressed size 641.84 MiB, 23.4% [6m53]
xz -7: total compressed size 621.82 MiB, 22.6% [7m19]
xz -8: total compressed size 599.33 MiB, 21.8% [7m18]
xz -9: total compressed size 588.18 MiB, 21.4% [7m40]
Note that compression takes much longer. More importantly, however,
decompression is much faster:
bzip2: 1m47.274s
xz -6: 0m55.446s
xz -7: 0m54.119s
xz -8: 0m52.388s
xz -9: 0m51.842s
The only downside to using -9 is that decompression takes a fair
amount (~65 MB) of memory.
Manifests are a huge pain, since users need to run nix-pull directly
or indirectly to obtain them. They tend to be large and lag behind
the available binaries; also, the downloaded manifests in
/nix/var/nix/manifest need to be in sync with the Nixpkgs sources. So
we want to get rid of them.
The idea of manifest-free operation works as follows. Nix is
configured with a set of URIs of binary caches, e.g.
http://nixos.org/binary-cache
Whenever Nix needs a store path X, it checks each binary cache for the
existence of a file <CACHE-URI>/<SHA-256 hash of X>.narinfo, e.g.
http://nixos.org/binary-cache/bi1gh9...ia17.narinfo
The .narinfo file contains the necessary information about the store
path that was formerly kept in the manifest, i.e., (relative) URI of
the compressed NAR, references, size, hash, etc. For example:
StorePath: /nix/store/xqp4l88cr9bxv01jinkz861mnc9p7qfi-neon-0.29.6
URL: 1bjxbg52l32wj8ww47sw9f4qz0r8n5vs71l93lcbgk2506v3cpfd.nar.bz2
CompressedHash: sha256:1bjxbg52l32wj8ww47sw9f4qz0r8n5vs71l93lcbgk2506v3cpfd
CompressedSize: 202542
NarHash: sha256:1af26536781e6134ab84201b33408759fc59b36cc5530f57c0663f67b588e15f
NarSize: 700440
References: 043zrsanirjh8nbc5vqpjn93hhrf107f-bash-4.2-p24 cj7a81wsm1ijwwpkks3725661h3263p5-glibc-2.13 ...
Deriver: 4idz1bgi58h3pazxr3akrw4fsr6zrf3r-neon-0.29.6.drv
System: x86_64-linux
Nix then knows that it needs to download
http://nixos.org/binary-cache/1bjxbg52l32wj8ww47sw9f4qz0r8n5vs71l93lcbgk2506v3cpfd.nar.bz2
to substitute the store path.
Note that the store directory is omitted from the References and
Deriver fields to save space, and that the URL can be relative to the
binary cache prefix.
This patch just makes nix-push create binary caches in this format.
The next step is to make a substituter that supports them.
In a private PID namespace, processes have PIDs that are separate from
the rest of the system. The initial child gets PID 1. Processes in
the chroot cannot see processes outside of the chroot. This improves
isolation between builds. However, processes on the outside can see
processes in the chroot and send signals to them (if they have
appropriate rights).
Since the builder gets PID 1, it serves as the reaper for zombies in
the chroot. This might turn out to be a problem. In that case we'll
need to have a small PID 1 process that sits in a loop calling wait().
In chroot builds, set the host name to "localhost" and the domain name
to "(none)" (the latter being the kernel's default). This improves
determinism a bit further.
P.S. I have to idea what UTS stands for.
This improves isolation a bit further, and it's just one extra flag in
the unshare() call.
P.S. It would be very cool to use CLONE_NEWPID (to put the builder in
a private PID namespace) as well, but that's slightly more risky since
having a builder start as PID 1 may cause problems.
On Linux it's possible to run a process in its own network namespace,
meaning that it gets its own set of network interfaces, disjunct from
the rest of the system. We use this to completely remove network
access to chroot builds, except that they get a private loopback
interface. This means that:
- Builders cannot connect to the outside network or to other processes
on the same machine, except processes within the same build.
- Vice versa, other processes cannot connect to processes in a chroot
build, and open ports/connections do not show up in "netstat".
- If two concurrent builders try to listen on the same port (e.g. as
part of a test), they no longer conflict with each other.
This was inspired by the "PrivateNetwork" flag in systemd.
Systemd can start the Nix daemon on demand when the Nix daemon socket
is first accessed. This is signalled through the LISTEN_FDS
environment variable, so all we need to do is check for that and then
use file descriptor 3 as the listen socket instead of creating one
ourselves.
We can't open a SQLite database if the disk is full. Since this
prevents the garbage collector from running when it's most needed, we
reserve some dummy space that we can free just before doing a garbage
collection. This actually revives some old code from the Berkeley DB
days.
Fixes#27.
For several platforms we don't currently have "native" Nix packages
(e.g. Mac OS X and FreeBSD). This provides the next best thing: a
tarball containing the closure of Nix, plus a simple script
"nix-finish-install" that initialises the Nix database, registers the
paths in the closure as valid, and runs "nix-env -i /path/to/nix" to
initialise the user profile.
The tarball must be unpacked in the root directory. It creates
/nix/store/... and /usr/bin/nix-finish-install. Typical installation
is as follows:
$ cd /
$ tar xvf /path/to/nix-1.1pre1234_abcdef-x86_64-linux.tar.bz2
$ nix-finish-install
(if necessary add ~/.nix-profile/etc/profile.d/nix.sh to the shell
login scripts)
After this, /usr/bin/nix-finish-install can be deleted, if desired.
The downside to the binary tarball is that it's pretty big (~55 MiB
for x86_64-linux).
