Monday, 2023-07-31

*** Dessira_ is now known as Dessira11:32
fricklerwell debian seems to have part of their builders hosted by osuosl, maybe Ramereth[m] is involved or knows someone? I might also have contacts at manda, the second host that is mentioned11:55
fungiyeah, i mean we have contacts and can probably find organizations interested in funding things, mostly just worth keeping an eye on for now but great to see it's progressing to an official architecture in debian finally13:05
fungiand there might be interest in things like like pyca/cryptography wheels for it, similar to how we help them out with arm64 builds13:06
Ramerethfungi Clark[m] frickler: riscv is still an emerging platform, especially for the datacenter space. We ended up using SiFive Unmatched boards and putting them in SuperMicro 1U cases, but it doesn't look pretty:
RamerethIn most cases, QEMU emulation on x86 is actually faster than current native hardware. But that might be changing soon. I was on a call last week with SiFive on how to get these in a better position for the data center (i.e. out-of-band, more cores, more memory, etc)15:22
RamerethWe have several projects that are just doing builds via qemu on x8615:22
fricklerRamereth: interesting, those pics indeed still look pretty adventurous15:32
Ramerethserial is only accessible via USB and we had to add a special cable to the power switch so that we could bounce the system via the PDU remotely if needed15:33
Ramerethas you can see, we daisy chained the usb so they could access it from other systems15:34
fricklernot sure if some pi with kvm hat could help. but that might then have more cpu power than the server itself :-D15:45
tonybI noticed: which I think we can remove given:
tonybbut maybe not given the org difference (opendev vs openstack)17:47
fungitonyb: i think it will work. note that our global zuul configuration still lives in openstack/project-config (site vars, tenants, et cetera)17:52
fungibut those are just files that get deployed to the servers17:52
fungithey're not used speculatively or anything17:52
fungithough that does mean we need to add them to site-variables.yaml first, make sure that gets deployed to the zuul servers, and then we can remove the old equivalents from the job config17:53
fungiif you want to propose changes for that i'm happy to review them17:53
tonybOkay.  That's a little more complex than I was thinking.  I'll do more research.17:57
funginot really all that complex at all. the site-variables.yaml gets automatically deployed whenever we merge a change to it18:06
fungishould be able to just add zuul_traceroute_host and zuul_image_manifest in there, then once it merges and the deploy job completes successfully we can test removing it from base-minimal18:07
fungitonyb: oh, i see what you mean, it's already set in site-variables.yaml18:10
fungiso no need to add it there18:11
tonybYeah as it's already there, although I notice that base-minimal uses: zuul_image_manifest but the site-variables supplies: zuul_site_image_manifest_files  so I'd need to reserach how to handle that18:12
fungiand it's already not in base or base-test, so we simply missed cleaning it up from base-minimal i think18:13
tonybThat's what I was thinking, As I was poking around it looked like an old comment and variables that can just be removed18:13
fungiyeah, i don't see any indication it was ever in base or base-test18:14
tonybI'll double check and push something up today18:14
fungiand looks like this is the only job that lists base-minimal as a parent:
fungiaccording to codesearch anyway18:15
fungiso i guess if we want to be ultra-careful we can trigger that job with a dnm change or something afterward18:16

Generated by 2.17.3 by Marius Gedminas - find it at!