Fork me on GitHub
#circleci
<
2021-01-07
>
borkdude15:01:24

What space do you get in circleci builds? > /usr/bin/ld: final link failed: No space left on device

timo16:01:22

Is there a good way to run a benchmark during CI on circleci? For Datahike we could use an instance that has a certain performance consistency to run a benchmark on.

borkdude16:01:03

@timok I've been told more than once that CI isn't a good environment to do this

borkdude16:01:17

unless you arrange your own bare metal probably

timo16:01:48

right. that is basically the requirement, to run a bare metal instance

conormcd16:01:48

https://circleci.com/docs/2.0/runner-overview/ is the only way I know of currently to run a CircleCI job on bare metal.

๐Ÿ‘ 3
glenjamin17:01:30

@timok using the machine executor is (currently) equivalent to getting an n2 instance on GCP, and will always be a dedicated VM

glenjamin17:01:35

so that might be sufficient for you

glenjamin17:01:28

the docker executor runs on a multi-tenant system, but we donโ€™t overprovision, so in theory it should be fairly consistent. In practice this varies

timo17:01:28

Cool, thanks! ๐Ÿ‘ I will probably start with a machine and take a look on consistency there. maybe improving it with something self-hosted later.

glenjamin17:01:37

i suspect the performance inconsistency is actually because we allow jobs to use spare CPU above their allocation if nothing else is using it, so if you land on a quiet machine you might go faster

๐Ÿ‘ 3