[09:48] <cjohnston> lol.. she has Ambers sweater too
[09:48] <cjohnston> uggh
[11:33] <nandersson> Hi, I saw the recording of Thomas Bushnell, Google on UDS. He made a strange remark. He said a reboot cost Google 1M USD - because they are tens of thousands of users and a reboot could take 15 min. This sounds strange. Does anybody know if they run their machines from NFS or something like that on Google???
[11:35] <zyga> nandersson, maybe it means that it takes 15 minutes to reboot to a state you were in before
[11:39] <nandersson> zyga, I dont know. He also said "A logout is only half of that". It seemed to me that everyone was connected to the same server via NFS or something like that.
[11:39] <zyga> nandersson, again, perhaps that's the cost of going back to the all the open applications you are looking at
[11:40] <PatrickDK> a reboot of what?
[11:40] <PatrickDK> if a server, could be nfs, or http server, or many other things
[11:40] <PatrickDK> but that sounds like you don't have a failover server setup, otherwise it would be a max, of a min, not 15
[11:41] <nandersson> PatrickDK, reboot of the users operating system (Goobuntu). Yeah, 15 min seems like a very long time. That is why I thought they might boot their machines from NFS or something.
[11:42] <PatrickDK> still doesn't matter, that is what failover is for
[11:42] <PatrickDK> I do iscsi boot, no local disks, and let that failover if I reboot the server
[11:43] <nandersson> PatrickDK, Yeah, it could be that Zyga is right. He might refer to boot back to a former state.
[11:43] <PatrickDK> sounds just completely nuts, if it costs you 1million to reboot, why not have a hot standby? that has got to be cheaper than 1million, for it's life
[11:43] <zyga> PatrickDK, hot standby on my current desktop state?
[11:43] <PatrickDK> yes, but if your booting back to a former state, and another server is handling the load
[11:43] <PatrickDK> how exactly could that add up to 1million?
[11:43] <zyga> and especially
[11:44] <zyga> I don't believe they meant servers
[11:44] <PatrickDK> oh for a user desktop?
[11:44] <zyga> the talk was about the desktop
[11:44] <PatrickDK> how does that affect thousands of users?
[11:44] <nandersson> Yeah, it didnt make any sense to me either. The talk was very interesting though. http://www.youtube.com/watch?feature=player_embedded&v=fu3pT_9nb8o#!
[11:44] <nandersson> Yeah, it was about the desktop
[11:57] <nandersson> In the same talk Thomas Busnell made a lot of remarks on Ubuntu Automated Testing, but I wonder - does it make sense for Google to test to compile the source code at all times to see if it breaks, dependencies fail, or what is that purpose? I guess Google would mainly be interested in seeing that the binaries work, and doesnt break any dependencies, that upgrades works, etcetera. Are those functions incorporated in Ubuntu Automated Testing?
[12:45] <patdk-wk> oh
[12:45] <patdk-wk> he is talking total
[12:46] <patdk-wk> 15min per user * thousands of desktops = 1Million cost
[12:46] <patdk-wk> not 1million per desktop
[12:46] <patdk-wk> or per reboot
[18:45] <stgraber> please ignore the new desktop images, they were just a mistake. 20120604.2 is still the one that needs testing
[19:07] <balloons> stgraber, whew.. :-) I was wondering
[19:07] <stgraber> balloons: yeah, both slangasek and I were building the same images ;)
[19:08] <balloons> stgraber, ohh, btw..  i guess we normally kill the dailies off during milestones, but now I can't get my reporting info for last week :-( Guess I need that db access
[19:08] <stgraber> balloons: I can mark them as "released" if you want, so they show up but are read-only
[19:09] <balloons> well, I mean, the results are hidden from me and everyone else.. it's an interesting conundrum. I don't want to confuse people
[19:14] <balloons> i've added it to my notes stgraber.. we'll discuss post-alpha 1
[19:15] <stgraber> balloons: actually, can you pastebin that script (I don't seem to have it around anymore), I think it should be easy to have it look a the archived milestones too
[19:17] <balloons> stgraber, d'oh! i'm too scatterbrained today.. lololololol
[19:17] <balloons> I was thinking it WASN'T in the archived runs
[19:18] <balloons> hence, my issues with not being able to get at stuff
[19:19] <balloons> I can pastebin the script anyways, if you'd like.. i started modifying it a bit, and then as we spoke about, decided it would be better to expend efforts towards getting a report into the interface itself