text stringlengths 0 2.68k |
|---|
**Dave Pacheco:** Yeah. So we've got to tell the operator "Look, you either need to do (what's called) expunge the sled", which means to remove it from the control plane, and we'll pretend like it's just caught fire; like, it's failed, and we've moved everything else elsewhere. Or you've got to figure out what's wrong ... |
**Jerod Santo:** Now, what did you think I was going to say? |
**Dave Pacheco:** I thought you were going to say an update that explodes; that like you start doing the update -- |
**Jerod Santo:** That was my next question. |
**Dave Pacheco:** ...and the control plane is now down, and - what do you do? |
**Jerod Santo:** Right. |
**Dave Pacheco:** And that's my nightmare. That's been my fear for the last couple of years. |
**Jerod Santo:** \[laughs\] That's why it's taken two years and you not done yet...? |
**Dave Pacheco:** Yeah, but seriously, that is why we've spent so much time on having the automation take these careful steps where every one of these steps we know is safe. As an example, we've got a CockroachDB cluster that's storing all the control plane database data... We've got five nodes. We definitely don't wan... |
**Jerod Santo:** Sure. |
**Dave Pacheco:** But it's a hard problem really, because part of what is involved in an upgrade is making backwards-incompatible changes to data formats, like database schemas, and things like that. And once you've done that, the old software can't read the new thing. So rollback is really not possible. What a lot of ... |
\[43:57\] But even then, there's still risk there, because whatever it is you're activating by taking that last step hasn't been tested before that. And there's kind of no way to get around that. And that's kind of a future problem for us right now, but it's something we're going to have to deal with. |
**Adam Stacoviak:** So it is your nightmare... Has it ever been, your nightmare? Meaning it's happened. |
**Dave Pacheco:** Like, in my career, yeah. |
**Adam Stacoviak:** Well, specifically for Oxide, and updates... |
**Dave Pacheco:** No, no. It hasn't. |
**Adam Stacoviak:** Any updates gone wrong? |
**Dave Pacheco:** No, but we haven't started doing that in production yet, so there's still time for that, I guess... |
**Adam Stacoviak:** How do you do it in production now? |
**Dave Pacheco:** The process we used -- |
**Jerod Santo:** The manual way... |
**Dave Pacheco:** Yeah, the manual thing when we -- so how the manual process that I was talking about at the very beginning works, basically it shuts down the whole control plane, and replaces all the software and brings it back up again. |
**Adam Stacoviak:** How long is that process? |
**Dave Pacheco:** It's actually not that -- it's shorter than the other way. |
**Adam Stacoviak:** I feel like that's safer. |
**Dave Pacheco:** It is safer. |
**Adam Stacoviak:** I mean, I get the whole reason -- |
**Jerod Santo:** Downtime, man. Can't have it. |
**Dave Pacheco:** It's downtime, and it's the self-service aspect. Although you could imagine a self-service version that looked more like that, but then the thing is like, if it's self-service, you're talking to an API, what's running that API while the thing is down? There's nothing, right? So that's why you've gotta... |
**Jerod Santo:** It's gotta be done. |
**Adam Stacoviak:** So downtime's required now. Future is -- you said non-destructive? |
**Jerod Santo:** Non-disruptive. |
**Dave Pacheco:** Non-disruptive, yeah. |
**Adam Stacoviak:** Meaning the VMs get migrated around versus shutdown, but you still reboot sleds. |
**Dave Pacheco:** That's right. |
**Adam Stacoviak:** You're still rebooting control planes, and stuff like that. |
**Dave Pacheco:** Yup. But that's not visible to them, because we have enough -- |
**Jerod Santo:** Is there a world in which that isn't even required? |
**Dave Pacheco:** Which part? |
**Jerod Santo:** What if you didn't have to reboot anything? |
**Dave Pacheco:** Oh. I think that's pretty dicey. It's definitely a thing that people have done... |
**Adam Stacoviak:** Hot swap? |
**Dave Pacheco:** Yeah. There's types of updates, or patches, running the latest kernel, where you kind of write the new one over here, and then you jump over there... But then the state -- |
**Jerod Santo:** It's just a pointer... \[laughs\] |
**Dave Pacheco:** It is. It's all software, right? |
**Jerod Santo:** That's right. Like, who says we have to reboot, you know? |
**Dave Pacheco:** Yeah. That is definitely harder to do with stuff like \[unintelligible 00:46:02.02\] job is to attest to the software that's currently running. |
**Jerod Santo:** Yeah. If you're going to change it out from underneath it \[unintelligible 00:46:11.13\] |
**Dave Pacheco:** Yeah. You need to figure out what that means. And then there's also this risk that you're now in a different state than you would have been in if you had actually bounced that thing. So have you created a time bomb for yourself, where if that thing loses power and actually does power back on, is it go... |
**Jerod Santo:** A catastrophic failure. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.