Page MenuHomePhabricator

Investigate and document cloudvirt-wdqs servers
Closed, ResolvedPublic

Description

I have IRC notifications for the string "wdqs" , and I noticed that 3 cloudvirt-wdqs alerted today. Per conversation with @Gehel , "the wikimedia-cloud team should manage the cloudvirt* servers" , "cloudvirt-wdqs are the physical servers hosting our WDQS VMs in WMCS" and "We don't have any active testing on WMCS for wikidata at the moment".

Creating this ticket to investigate and document current and past use for cloudvirt-wdqs hosts, so we avoid any future confusion about who owns them, if we still need them, etc...

Event Timeline

Hello bking! Those hosts were bought as a result of T221631

They were purchased specifically to host great big wdqs-specific VMS in cloud-vps. They've been used for that purpose off and on. The hardware is definitely maintained by the WMCS team, as part of the cloudvirt fleet.

That said -- they've been idle for quite some time, waiting for the wdqs people to request a new workload there. If you don't anticipate needing them any longer then we might repurpose the hardware for other uses. The hardware itself is scheduled for refresh in 2024.

Oh, I should add: they alerted the other day because I failed to catch them with a wildcard when doing some routine maintenance. I did my best to silence the alerts but they managed to get forwarded anyway, which is a mystery relegated to T324208

Hey Andrew, thanks for the follow-up!

I'm open to letting WMCS repurpose the hardware, but let me confirm with my team first. Will update here or ping in IRC when I have more info.

Hello Andrew,

After discussing it with my team, we still want to hold on to these servers for now. We still need to completely replace the most important application in the WDQS stack (Blazegraph) , and we believe we'll need the capacity to test at scale.

Sorry for the confusion and let us know if you have any other questions.

That's just fine -- as long as those servers are good for something we'll keep them around! lmk when you're ready to stand up VMs there.

Gehel claimed this task.