Commit graph

46 commits

Author SHA1 Message Date
Eelco Dolstra 7863d2e1da Step cancellation: Don't use pthread_cancel()
This was a bad idea because pthread_cancel() is unsalvageable broken
in C++. Destructors are not allowed to throw exceptions (especially in
C++11), but pthread_cancel() can cause a __cxxabiv1::__forced_unwind
exception inside any destructor that invokes a cancellation
point. (This exception can be caught but *must* be rethrown.) So let's
just kill the builder process instead.
2016-11-07 19:38:24 +01:00
Eelco Dolstra 3b84d4711b Bump Nix 2016-10-26 15:10:56 +02:00
Eelco Dolstra 8e1d791d0c Truncate the log just before starting the remote build
This gets rid of all those remote substitution messages that were
polluting the build logs.
2016-10-26 13:41:51 +02:00
Eelco Dolstra 3fcfa20d1a Fix regression caused by ee2e9f53
‘basicDrv.inputSrcs’ also contains the outputs of inputDrvs. These
don't necessarily exist in the local store, so copying them may cause
an exception. We should only copy the real inputSrcs.
2016-10-24 16:49:11 +02:00
Eelco Dolstra ee2e9f5335 Update to reflect BinaryCacheStore changes
BinaryCacheStore no longer implements buildPaths() and ensurePath(),
so we need to use copyPath() / copyClosure().
2016-10-07 20:23:05 +02:00
Eelco Dolstra 6a313c691b hydra-queue-runner: Fix build 2016-10-06 16:58:54 +02:00
Alexander Ried 1c2f6281b9 Remove signing parameter (nix#f435f82) 2016-10-06 15:09:12 +02:00
Eelco Dolstra b1512a152a Fix build failure on GCC 5.4 2016-09-30 17:05:07 +02:00
Eelco Dolstra ad834343b5 Fix build against current Nix master 2016-04-13 16:30:52 +02:00
Eelco Dolstra ddc9f3cc6a Temporarily disable machines on any exception, not just connection failures 2016-03-22 16:54:40 +01:00
Eelco Dolstra 4151be7e69 Make the output size limit configurable
The maximum output size per build step (as the sum of the NARs of each
output) can be set via hydra.conf, e.g.

  max-output-size = 1000000000

The default is 2 GiB.

Also refactored the build error / status handling a bit.
2016-03-09 17:00:09 +01:00
Eelco Dolstra 9127f5bbc3 hydra-queue-runner: Limit memory usage
When using a binary cache store, the queue runner receives NARs from
the build machines, compresses them, and uploads them to the
cache. However, keeping multiple large NARs in memory can cause the
queue runner to run out of memory. This can happen for instance when
it's processing multiple ISO images concurrently.

The fix is to use a TokenServer to prevent the builder threads to
store more than a certain total size of NARs concurrently (at the
moment, this is hard-coded at 4 GiB). Builder threads that cause the
limit to be exceeded will block until other threads have finished.

The 4 GiB limit does not include certain other allocations, such as
for xz compression or for FSAccessor::readFile(). But since these are
unlikely to be more than the size of the NARs and hydra.nixos.org has
32 GiB RAM, it should be fine.
2016-03-09 14:30:13 +01:00
Eelco Dolstra 922dc541c2 Add log message 2016-02-29 11:58:06 +01:00
Eelco Dolstra 6bb860fd6e Add FIXME 2016-02-26 21:15:05 +01:00
Eelco Dolstra b9afaadfb3 Keep better bytesReceived/bytesSent stats 2016-02-26 16:17:05 +01:00
Eelco Dolstra 6d741d2ffa Prevent download of NARs we just uploaded 2016-02-26 15:21:44 +01:00
Eelco Dolstra db3fcc0f5e Enable substitution on the build machines
If properly configured, this allows them to get store paths directly
from S3, rather than having to receive them from the queue runner.
2016-02-18 16:42:05 +01:00
Eelco Dolstra ce5790285a Merge remote-tracking branch 'origin/master' into binary-cache 2016-02-17 11:54:59 +01:00
Eelco Dolstra d7a123fcd4 Keep track of the time we spend copying to/from build machines 2016-02-17 10:30:23 +01:00
Eelco Dolstra 2d0dd7fb49 hydra-queue-runner: Write directly to a binary cache 2016-02-15 21:10:29 +01:00
Eelco Dolstra 92d8b59361 Process Nix API changes 2016-02-11 15:59:47 +01:00
Eelco Dolstra 8e8e31ce86 Re-implement log size limits
The old queue runner already had this. However, we now store "log
limit exceeded" as a separate status code in the database.
2015-10-06 17:35:08 +02:00
Eelco Dolstra 6075ac6fed Remove localhost hack 2015-09-09 16:50:59 +02:00
Eelco Dolstra 2a7fbd57cc Allow the machines file to specify host public keys
It's easier for the Hydra provisioner to put host public keys in the
machines file than to separately manage the known_hosts file
(especially when the provisioner runs on a different machine).
2015-08-26 13:43:02 +02:00
Eelco Dolstra ff3f5eb4d8 Fix remote building on Nix 1.10 2015-07-31 03:41:55 +02:00
Eelco Dolstra 5b9a288123 Workaround for RemoteStore not supporting cmdBuildDerivation yet 2015-07-31 03:39:20 +02:00
Eelco Dolstra c18fb0ad74 Temporarily disable machines after a connection failure 2015-07-21 15:58:47 +02:00
Eelco Dolstra 5370be9f52 hydra-queue-runner: Use cmdBuildDerivation
See 1511aa9f48 and eda2f36c2a.
2015-07-21 01:54:24 +02:00
Eelco Dolstra 3ded87329d Keep track of how many threads are waiting 2015-07-10 19:10:14 +02:00
Eelco Dolstra 35b7c4f82b Allow only 1 thread to send a closure to a given machine at the same time
This prevents a race where multiple threads see that machine X is
missing path P, and start sending it concurrently. Nix handles this
correctly, but it's still wasteful (especially for the case where P ==
GHC).

A more refined scheme would be to have per machine, per path locks.
2015-07-07 14:06:48 +02:00
Eelco Dolstra 63745b8e25 Move buildRemote() into State 2015-07-07 10:25:33 +02:00
Eelco Dolstra c6fcce3b3b Moar stats 2015-06-25 16:47:39 +02:00
Eelco Dolstra 18a3c3ff1c Update "make check" for the new queue runner
Also, if the machines file contains an entry for localhost, then run
"nix-store --serve" directly, without going through SSH.
2015-06-25 16:47:39 +02:00
Eelco Dolstra 1a0e1eb5a0 More stats 2015-06-24 13:19:27 +02:00
Eelco Dolstra 681f63a382 Typo 2015-06-23 02:15:11 +02:00
Eelco Dolstra 4db7c51b5c Rate-limit the number of threads copying closures at the same time
Having a hundred threads doing I/O at the same time is bad on magnetic
disks because of the excessive disk seeks. So allow only 4 threads to
copy closures in parallel.
2015-06-23 01:49:14 +02:00
Eelco Dolstra 44a2b74f5a Keep track of the number of build steps that are being built
(As opposed to being in the closure copying stage.)
2015-06-22 11:23:00 +02:00
Eelco Dolstra 133d298e26 Asynchronously compress build logs 2015-06-19 15:06:12 +02:00
Eelco Dolstra 3855131185 hydra-queue-runner: Improve SSH flags 2015-06-18 00:50:48 +02:00
Eelco Dolstra f57d0b0c54 hydra-queue-runner: Maintain count of active build steps 2015-06-18 00:24:56 +02:00
Eelco Dolstra ca48818b30 Fix remote building 2015-06-17 17:28:59 +02:00
Eelco Dolstra 745efce828 hydra-queue-runner: Implement timeouts
Also, keep track of timeouts in the database as a distinct build
status.
2015-06-17 13:32:33 +02:00
Eelco Dolstra b91a616520 Automatically retry aborted builds
Aborted builds are now put back on the runnable queue and retried
after a certain time interval (currently 60 seconds for the first
retry, then tripled on each subsequent retry).
2015-06-17 11:45:20 +02:00
Eelco Dolstra c68036f8b0 Pass ssh key 2015-06-10 14:57:07 +02:00
Eelco Dolstra 7dd1f0097e Finish copyClosure 2015-06-09 16:03:41 +02:00
Eelco Dolstra 8b12ac1f6d Basic remote building
This removes the need for Nix's build-remote.pl.

Build logs are now written to $HYDRA_DATA/build-logs because
hydra-queue-runner doesn't have write permission to /nix/var/log.
2015-06-09 14:21:21 +02:00