Re: ovirt-engine-nodejs-modules build-artifacts failing on el7 and fc30
The nexus instance should be available for everyone:
https://nexus.apps.ovirt.org/repository/npm/
Is it possible to set several repos in yarn so that it can access upstream
directly if nexus is not reachable for some reason?
Alternatively, can we just pass a custom repo in pur CI as an env var or
just create it for a run that runs in our CI so that nexus is used only
when a certain condition is met?
A quick google search shows that yarn cannot handle multiple repos by
default: https://github.com/yarnpkg/yarn/issues/547
I am not sure however if there are common workarounds for this in practise
that could allow us to use nexus and fail over if it's not reachable.
On Mon, Mar 9, 2020 at 10:12 AM Michal Skrivanek <
michal.skrivanek(a)redhat.com> wrote:
> On 9 Mar 2020, at 09:09, Anton Marchukov <amarchuk(a)redhat.com> wrote:
>
> Hello Scott.
>
>> On 6 Mar 2020, at 22:23, Scott Dickerson <sdickers(a)redhat.com> wrote:
>>
>> Advice ranges from increasing yarn timeouts to decreasing the size of
packages uploaded to their registry so caching stuff on the proxy and not
relying on upstream seems like the way to go.
>>
>>
>> Caching on the proxy won't work since everything is https, unless squid
is setup to intercept https traffic...
>
> Just to remind that in the past we have set up a Nexus server in oVirt
PHX DC that is capable of caching nodejs artefacts. AFAIK it is not
actively used, but does it makes sense to reconsider this decisions now? I
think using Nexus is more clear that doing SSL MITM on existing squid
(although this is also something doable).
But we can’t easily just change the urls in upstream code unless that
server is accessible externally
>
> --
> Anton Marchukov
> Associate Manager - RHV DevOps - Red Hat
>
>
>
>
>
>
--
Regards,
Evgheni Dereveanchin
Attachments:
- attachment.html (text/html — 2.7 KB)