Commit graph

261 commits

Author SHA1 Message Date
Eelco Dolstra
faf87b51f7
Show why GC socket connection was refused
Co-authored-by: John Ericson <git@JohnEricson.me>
2024-01-08 14:14:36 +01:00
Eelco Dolstra
c5fdbdae32 LocalStore::addTempRoot(): Handle ENOENT
If the garbage collector has acquired the global GC lock, but hasn't
created the GC socket yet, then a client attempting to connect would
get ENOENT. Note that this only happens when the GC runs for the first
time on a machine. Subsequently clients will get ECONNREFUSED which
was already handled.

Fixes #7370.
2023-06-20 11:14:55 +02:00
Eelco Dolstra
4f3a4b732c Make findRuntimeRoots() more resilient to disappearing processes
I saw this random failure in https://hydra.nixos.org/build/211811692:

  error: opening /proc/15307/fd: No such process

while running nix-collect-garbage in a readfile-context.sh. This is
because we're not handling ESRCH errors reading /proc/<pid>/fd. So
just move the read inside the try/catch where we do handle it.
2023-03-09 16:44:51 +01:00
Eelco Dolstra
29abc8e764 Remove FormatOrString and remaining uses of format() 2023-03-02 15:57:54 +01:00
Eelco Dolstra
28d5b5cd45 Fix deadlock between auto-GC and addTempRoot()
Previously addTempRoot() acquired the LocalStore state lock and waited
for the garbage collector to reply. If the garbage collector is in the
same process (as it the case with auto-GC), this would deadlock as
soon as the garbage collector thread needs the LocalStore state lock.

So now addTempRoot() uses separate Syncs for the state that it
needs. As long at the auto-GC thread doesn't call addTempRoot() (which
it shouldn't), it shouldn't deadlock.

Fixes #3224.
2023-01-03 15:20:21 +01:00
Eelco Dolstra
224b56f10e Move creation of the temp roots file into its own function
This also moves the file handle into its own Sync object so we're not
holding the _state while acquiring the file lock. There was no real
deadlock risk here since locking a newly created file cannot block,
but it's still a bit nicer.
2023-01-03 14:51:23 +01:00
Eelco Dolstra
5b798f6cae Fix random client failures during GC server shutdown
We need to close the GC server socket before shutting down the active
GC client connections, otherwise a client may (re)connect and get
ECONNRESET. But also handle ECONNRESET for resilience.

Fixes random failures like

  GC socket disconnected
  connecting to '/tmp/nix-shell.y07M0H/nix-test/default/var/nix/gc-socket/socket'
  sending GC root '/tmp/nix-shell.y07M0H/nix-test/default/store/kb5yzija0f1x5xkqkgclrdzldxj6nnc6-non-blocking'
  reading GC root from client: error: unexpected EOF reading a line
  1 store paths deleted, 0.00 MiB freed
  error: reading from file: Connection reset by peer

in gc-non-blocking.sh.
2022-11-27 12:57:18 +01:00
Andrew Brooks
565d888e0f Address PR feedback on #6694 2022-09-12 11:33:23 -05:00
Andrew Brooks
84fe75a12a Keep created temp dirs inside store, but protect from GC
Implements the approach suggested by feedback on PR #6994, where
tempdir paths are created in the store (now with an exclusive lock).

As part of this work, the currently-broken and unused
`createTempDirInStore` function is updated to create an exclusive lock
on the temp directory in the store.

The GC now makes a non-blocking attempt to lock any store directories
that "look like" the temp directories created by this function, and if
it can't acquire one, ignores the directory.
2022-09-06 17:48:00 -05:00
Théophane Hufschmitt
d71d9e9fbf moveFile -> renameFile
`move` tends to have this `mv` connotation of “I will copy it for you if
needs be”
2022-08-03 10:27:25 +02:00
Théophane Hufschmitt
c2de0a232c Create a wrapper around stdlib’s rename
Directly takes some c++ strings, and gently throws an exception on error
(rather than having to inline this logic everywhere)
2022-08-03 10:27:25 +02:00
Anders Kaseorg
754cd53faf Add missing rethrows in conditional exception handlers
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
2022-06-10 10:52:41 -07:00
Sergei Trofimovich
d58453f72e gc: don't visit implicit referrers on garbage collection
Before the change garbage collector was not considering
`.drv` and outputs as alive even if configuration says otherwise.
As a result `nix store gc --dry-run` could visit (and parse)
`.drv` files multiple times (worst case it's quadratic).

It happens because `alive` set was populating only runtime closure
without regard for actual configuration. The change fixes it.

Benchmark: my system has about 139MB, 40K `.drv` files.

Performance before the change:

    $ time nix store gc --dry-run
    real    4m22,148s

Performance after the change:

    $ time nix store gc --dry-run
    real    0m14,178s
2022-03-17 18:47:29 +00:00
Sergei Trofimovich
6b1872312f nix store gc: account for auto-optimised store
Before the change on a system with `auto-optimise-store = true`:

    $ nix store gc --verbose --max 1

deleted all the paths instead of one path (we requested 1 byte limit).

It happens because every file in `auto-optimise-store = true` has at
least 2 links: file itself and a link in /nix/store/.links/ directory.

The change conservatively assumes that any file that has one (as before)
or two links (assume auto-potimise mode) will free space.

Co-authored-by: Sandro <sandro.jaeckel@gmail.com>
2022-03-13 07:24:48 +00:00
Eelco Dolstra
a949673a5b Fix Darwin build
Fixes #6169
2022-02-28 15:48:26 +01:00
Eelco Dolstra
df552ff53e Remove std::string alias (for real this time)
Also use std::string_view in a few more places.
2022-02-25 16:13:02 +01:00
Eelco Dolstra
d17d46cfc2 More debug info 2021-12-13 19:03:08 +01:00
Eelco Dolstra
05fcc681ac Retry on ECONNREFUSED
https://hydra.nixos.org/build/161439235
2021-12-13 19:00:19 +01:00
Eelco Dolstra
6d30f9e6fe Explicitly make GC roots client sockets blocking
On macOS / BSD, these sockets inherit the non-blocking flag of the
server soocket, which is not what we want.

https://github.com/dotnet/runtime/issues/25069
https://bugs.python.org/issue7995

https://hydra.nixos.org/build/161439304
2021-12-13 18:59:50 +01:00
Alex Shabalin
db2e4489a5 Unify #if linux 2021-11-22 13:57:56 +01:00
Alex Shabalin
a18d9269a5 Fix build warnings on MacOS 2021-11-19 15:22:31 +01:00
Eelco Dolstra
33d04e8a8d Use nix::connect() to connect to the garbage collector 2021-10-28 22:51:11 +02:00
Eelco Dolstra
22c35ea5b8 Remove unused variable 2021-10-28 22:51:11 +02:00
Eelco Dolstra
10f9a8e77d Add a test for the non-blocking GC 2021-10-15 16:52:37 +02:00
Eelco Dolstra
ac54c6faa6 Fix main GC thread exiting 2021-10-15 16:36:48 +02:00
Eelco Dolstra
e0936ae38f Fix crash when a GC client disconnects
The client thread can't just delete its own thread object from
connections, it has to detach it.
2021-10-15 16:12:21 +02:00
Eelco Dolstra
304180d0de Memoize queryReferrers() 2021-10-15 12:20:23 +02:00
Eelco Dolstra
17e6ebcc90 Speed up GC by marking entire closures as live 2021-10-14 14:13:57 +02:00
Eelco Dolstra
0154fa30cf Remove GCState 2021-10-14 13:52:49 +02:00
Eelco Dolstra
0317ffdad3 Move deleteFromStore() 2021-10-14 13:34:48 +02:00
Eelco Dolstra
eab934cb2a Make the canReachRoots() traversal non-recursive 2021-10-14 12:34:32 +02:00
Eelco Dolstra
35c98a59c5 Fix GC when there are cycles in the referrers graph
(where "referrers" includes the reverse of derivation outputs and
derivers). Now we do a full traversal to look if we can reach any
root. If not, all paths reached can be deleted.
2021-10-13 12:12:44 +02:00
Eelco Dolstra
1785ba2980 Simplify 2021-10-13 12:12:44 +02:00
Eelco Dolstra
262520fcfe Use a thread per connection 2021-10-13 12:12:44 +02:00
Eelco Dolstra
ff453b06f9 Fix auto-gc 2021-10-13 12:12:44 +02:00
Eelco Dolstra
8614cf1334 Non-blocking garbage collector
The garbage collector no longer blocks other processes from
adding/building store paths or adding GC roots. To prevent the
collector from deleting store paths just added by another process,
processes need to connect to the garbage collector via a Unix domain
socket to register new temporary roots.
2021-10-13 12:12:44 +02:00
Eelco Dolstra
9947f1646a Remove syncWithGC() 2021-10-13 12:12:44 +02:00
Eelco Dolstra
8eac7dfad4 Remove trash directory 2021-10-13 12:12:44 +02:00
Eelco Dolstra
4202a3bc4e UDSRemoteStore: Support the 'root' store parameter
Useful when we're using a daemon with a chroot store, e.g.

  $ NIX_DAEMON_SOCKET_PATH=/tmp/chroot/nix/var/nix/daemon-socket/socket nix-daemon --store /tmp/chroot

Then the client can now connect with

  $ nix build --store unix:///tmp/chroot/nix/var/nix/daemon-socket/socket?root=/tmp/chroot nixpkgs#hello
2021-06-18 17:04:11 +02:00
volth
eee18f88dd
Handle amount of disk space saved by hard linking being negative
Fixes bogus messages like "currently hard linking saves 17592186044416.00 MiB".
2020-10-12 16:06:38 +00:00
John Ericson
6cc1541782 Split out local-fs-store.hh
This matches the already-existing `local-fs-store.cc`.
2020-10-09 20:18:08 +00:00
Eelco Dolstra
236d9ee7f7 lstat() cleanup 2020-09-23 19:17:28 +02:00
regnat
520895b1da Fix garbage collection of CA derivations
Fix #4026
2020-09-17 13:36:58 +02:00
Eelco Dolstra
8a945d6ddb
Remove gc-check-reachability 2020-09-03 11:30:15 +02:00
Eelco Dolstra
82b77a7726
addPermRoot(): Remove indirect flag 2020-09-03 11:29:34 +02:00
Eelco Dolstra
3f6e88a552 unsigned long long -> uint64_t 2020-07-30 13:34:04 +02:00
Eelco Dolstra
c0dd05131e toStorePath(): Return a StorePath and the suffix 2020-07-13 16:25:48 +02:00
Eelco Dolstra
29542865ce Remove StorePath::clone() and related functions 2020-06-16 22:20:18 +02:00
Eelco Dolstra
1fb762d11f Get rid of explicit ErrorInfo constructors 2020-06-15 14:06:58 +02:00
Ben Burdette
ef9dd9f9bc formatting and a few minor changes 2020-05-13 15:56:39 -06:00