This allows for better builder usage when the queue runner is busy. To
avoid running into uncontrollable imbalances between builder/queue
runner, we only release the machine reservation after the local
throttler has found a slot to start copying the outputs for that build.
Each output for a given step being ingested is looked up in parallel,
which should basically multiply the speed of builds ingestion by the
average number of outputs per derivation.
Running the query with/without it shows that it makes no difference to
postgres, since there's an index on finished=0 already. This allows a
few simplifications, but also paves the way towards running multiple
parallel monitor threads in the future.
By looking at the ratio of running vs. waiting for the dispatcher and
the queue monitor, we should get better visibility into what hydra is
currently bottlenecked on.
There are other side effects we can try to measure to get to the same
result, but having a simple way doesn't cost us much.
My current theory is that running more parallel xz than available CPU
cores is reducing our overall throughput by requiring more scheduling
overhead and more cache thrashing.
Closes#1336
When restarting postgresql, the connections are still reused in
`hydra-queue-runner` causing errors like this
main thread: Lost connection to the database server.
queue monitor: Lost connection to the database server.
and no more builds being processed.
`hydra-evaluator` doesn't have that issue since it crashes right away.
We could let it retry indefinitely as well (see below), but I don't
want to change too much.
If the DB is still unreachable 10s later, the process will stop with a
non-zero exit code because of a missing DB connection. This however
isn't such a big deal because it will be immediately restarted
afterwards. With the current configuration, Hydra will never give up,
but restart (and retry) infinitely. To me that seems reasonable, i.e. to
retry DB connections on a long-running process. If this doesn't work
out, the monitoring should fire anyways because the queue fills up, but
I'm open to discuss that.
Please note that this isn't reproducible with the DB and the queue
runner on the same machine when using `services.hydra-dev`, because of
the `Requires=` dependency `hydra-queue-runner.service` ->
`hydra-init.service` -> `postgresql.service` that causes the queue
runner to be restarted on `systemctl restart postgresql`.
Internally, Hydra uses Nix's pool data structure: it basically has N
slots (here DB connections) and whenever a new one is requested, an idle
slot is provided or a new one is created (when N slots are active, it'll
be waited until one slot is free). The issue in the code here is however
that whenever an error is encountered, the slot is released, however the
same broken connection will be reused the next time. By using
`Pool::Handle::markBad`, Nix will drop a broken slot. This is now being
done when `pqxx::broken_connection` was caught.
This is *just* using the fields from that type, and only where the types
coincide. (There are two fields with different types, `speedFactor` most
interestingly.) No code is reused, so we can be sure that no behavior is
changed.
Once the types are reconciled on the Nix side, then we can start
carefully actually reusing code.
Progress on #1164
- Use the type itself
This lays the foundation for being able to dedup the protocol code.
- Use `BasicConnection::handshake`, replacing ours.
- Use `BasicConnection::queryValidPaths`
- Use `BasicConnection::putBuildDerivationRequest`
This is just C++ changes without any Perl / Frontend / SQL Schema
changes.
The idea is that it should be possible to redeploy Hydra with these
chnages with (a) no schema migration and also (b) no regressions. We
should be able to much more safely deploy these to a staging server and
then production `hydra.nixos.org`.
Extracted from #875
Co-Authored-By: Théophane Hufschmitt <theophane.hufschmitt@tweag.io>
Co-Authored-By: Alexander Sosedkin <monk@unboiled.info>
Co-Authored-By: Andrea Ciceri <andrea.ciceri@autistici.org>
Co-Authored-By: Charlotte 🦝 Delenk Mlotte@chir.rs>
Co-Authored-By: Sandro Jäckel <sandro.jaeckel@gmail.com>
Periodically, I have seen tests fail because of out of order queue runner behavior:
checking the queue for builds > 0...
loading build 1 (tests:basic:empty_dir)
aborting unsupported build step '...-empty-dir.drv' (type 'x86_64-linux')
marking build 1 as failed
adding new machine ‘localhost’
This patch should prevent the dispatcher from running before any machines are
made available.
Previously, the build ID would never flow through channels which
exited.
This patch tracks the buildOne state as part of State and exits avoids
waiting forever for new work.
The code around buildOnly is a bit rough, making this a bit weird to
implement but since it is only used for testing the value of improving
it on its own is a bit questionable.
If we don't see machine that supports a build step for
'max_unsupported_time' seconds, the step is aborted. The default is 0,
which is appropriate for Hydra installations that don't provision
missing machines dynamically.
(cherry picked from commit f5cdbfe21d)
In a NixOS container, cmdBuildDerivation doesn't work because we're
not privileged. But we also don't need it because the store already
has the derivation.
Also, don't copy from/to the store since this gives errors about
missing signatures.
K900andMartin Weinelt
Pierre BourdonandMartin Weinelt
John Ericson
Maximilian Bosch
Jörg Thalheim
Graham ChristensenandGitHub
Cole Helbling
ajs124andGraham Christensen
Théophane Hufschmitt
Eelco Dolstra