Time |
Nick |
Message |
11:40 |
lfaraone |
basically, I asked if they had any spare servers in Sept, and they emailed today saying "hi, we've got them in sets of 3, we'll send you four boxes if you pick up shipping" |
11:40 |
bernie |
hello |
11:40 |
|
dogi is coming |
11:40 |
silbe |
#topic server donations from WikiMedia foundation |
11:40 |
lfaraone |
they've been in service for three year, and are out of warranty, which is why WMF doens't want them. |
11:41 |
|
will brb in less than 5. |
11:41 |
silbe |
oh, that was a quit, not a join... let's wait for him. |
11:42 |
dogi |
#startmeeting |
11:43 |
|
hui |
11:43 |
walterbender |
bernie: if you want them, we can cover the shipping, I am sure |
11:43 |
silbe |
#link http://meeting.sugarlabs.org/s[…]0100113_1140.html |
11:43 |
dogi |
hi all |
11:43 |
walterbender |
dogi: long time, no see :) |
11:43 |
dogi |
:) |
11:44 |
silbe |
dogi: lfaraone will join us in < 5 minutes |
11:45 |
bernie |
the specs are not impressive, but you don't look a gift server in the case |
11:45 |
lfaraone |
Back. |
11:46 |
silbe |
ok, let's start then |
11:46 |
dogi |
#link http://me.etin.gs/treehouse/tr[…]0100112_1546.html |
11:46 |
|
+1 |
11:46 |
bernie |
Dual CPU 2.5 GHz AMD |
11:46 |
|
* 3-4GB RAM Each |
11:46 |
|
* Most have 80 GB or larger HDD |
11:47 |
silbe |
we need to decide NOW, so I'd argue that we give more priority to the "we want them" side - we can always send them elsewhere later if it turns out we have no use for them |
11:48 |
|
they're not top-notch, but still pretty decent (likely better than anything i own) |
11:48 |
lfaraone |
So I can host 3U of them in Virginia. |
11:48 |
|
(one at TW IX, where we have a gur. 10mbits connection, one at the ACC, (as long as we don't have heavy network usage), and one at Obscure.org) |
11:48 |
dogi |
+1 silbe |
11:49 |
silbe |
one pro argument i've given yesterday is that we might use a set of them to experiment with scaling aslo - as (rightly) argued on systems@ it doesn't make a lot of sense to set up three VMs on the same host for that |
11:49 |
lfaraone |
Since they are shipped in groups of 3, I was figuring we can send 2 boxes (6) to Walter and the other two to the ACC, where we can move them to TWIX and Obscure. |
11:50 |
bernie |
we need to ask RIT how much rack space we're allowed to use |
11:50 |
silbe |
what's ACC and Obscure? I guess TWIX is the ISP mentioned yesterday... |
11:51 |
lfaraone |
silbe: ACC is the arlington career center, c/o Jeff. |
11:51 |
bernie |
silbe: we have an offer from transworldix to rack one machine there |
11:51 |
|
silbe: and one at rit |
11:51 |
silbe |
so ACC=RIT, TWIX=transworldix? |
11:51 |
lfaraone |
silbe: Obscure is a non-profit ISP-ish organization that does some sugar and educational stuff. |
11:51 |
|
silbe: RIT == rit.edu |
11:51 |
|
silbe: RIT is in NY |
11:51 |
dogi |
is interested in checking some them ... he is right now in boston :) |
11:51 |
bernie |
since the machines are 1U, perhaps racking 3 of them wouldn't be a problem for rit? |
11:52 |
dogi |
+1 bernie |
11:52 |
walterbender |
bernie: so maybe ship one box to Boston, one to NY? |
11:52 |
dogi |
who is NY? walter? |
11:52 |
lfaraone |
dogi: RIT is. |
11:52 |
|
dogi: and dfarning, iirc. |
11:53 |
dogi |
thx |
11:53 |
walterbender |
lfaraone: dfarning is in WI :) |
11:53 |
silbe |
yeah, let's dfarning take the aslo servers into his home ;) |
11:53 |
dogi |
+1 walterbender |
11:54 |
cjb |
bernie: (just read scrollback) looks good! |
11:54 |
bernie |
invites djbclark |
11:54 |
lfaraone |
okay, so the plan is 2 boxes (6 servers) to ACC in VA, 1b/3s to Walter in MA, and 1b/3s to RIT in NY, afaict. |
11:54 |
dogi |
+1 cjb |
11:55 |
lfaraone |
(two to ACC since we have multiple area partners) |
11:55 |
cjb |
whoa |
11:55 |
|
there are 12 servers? |
11:55 |
|
I don't think we need 12 servers |
11:56 |
silbe |
cjb: they are out of warranty; WMF recommended to use some of them as spares |
11:56 |
lfaraone |
cjb: WMF told us that they don't want to send less, since some may break. |
11:56 |
bernie |
cjb: yeah, I wouldn't take more than 3 |
11:56 |
cjb |
ah, interesting |
11:57 |
lfaraone |
cjb: they have been running Wikipedia for 3 years, it's every server's worst nightmare. |
11:57 |
dogi |
lol |
11:57 |
|
likes the idea wikipedia sponsoring us :P |
11:58 |
bernie |
if we're going to cluster aslo, we require minimum 2 machines in the same location. |
11:58 |
silbe |
the only backside of the deal is that they're rather low on HD space |
11:58 |
bernie |
otherwise, I see no reason to move it away from sunjammer |
11:58 |
|
silbe: aslo does not need much disk space |
11:58 |
lfaraone |
silbe: 80gb? |
11:59 |
silbe |
bernie: how much does it need, actually? |
11:59 |
|
lfaraone: that's not even enough for 8 buildslaves :-/ |
12:00 |
|
(we have bender for that, just using as a comparison) |
12:00 |
bernie |
silbe: lol |
12:00 |
dogi |
silbe, lol for messuring in buildslaves |
12:01 |
bernie |
silbe: since bender has very little b/w and had been not very reliable in the past, I wouldn't mind replacing it in the future |
12:01 |
|
silbe: current aslo is 4.5GB, plus the db |
12:01 |
silbe |
so we already have a use for some of them: aslo scaling and buildslaves |
12:02 |
bernie |
but we want the db centralized of course |
12:02 |
silbe |
bernie: ok, sounds reasonable enough. hey, we could even move it to RAM ;) |
12:03 |
lfaraone |
bernie: we can't do anything bandwidth-intensive at ACC, but we can with TWIX. (Obscure is in the middle somewhat on that) |
12:03 |
|
I'll see if I can squeeze another U or two out of TWIX> |
12:03 |
bernie |
another way to scale aslo up in the short term would be to purchase a modern server machine which would be 8-10 times faster than sunjammer |
12:03 |
|
this is the simplest solution and therefore my favorite |
12:03 |
|
it does not require special sysadmin skills |
12:04 |
|
it is much much more robust |
12:04 |
cjb |
bernie: could a WMF machine be good enough? |
12:04 |
|
(I agree that using one machine should be preferred) |
12:04 |
bernie |
cjb: we have found cheap hardware with 8 cores |
12:04 |
|
neuralis: ping? |
12:04 |
silbe |
bernie: but we still need to scale aslo to multiple machines in the future - we just move the breaking point (which is good, of course) |
12:05 |
|
djbclark: http://meeting.sugarlabs.org/s[…]0100113_1140.html |
12:06 |
djbclark |
silbe: thx |
12:06 |
bernie |
silbe: yes, this is only a short term solution |
12:06 |
cjb |
bernie: right, but that doesn't really answer the question |
12:06 |
|
bernie: how many cores do the WMF machines have? do you think they could handle ASLO? |
12:06 |
bernie |
silbe: "short term" probably means 2 years at the current growth rate |
12:07 |
dogi |
we have in the end to decide only if we want them first? |
12:07 |
bernie |
cjb: they don't say. probably one core |
12:07 |
dogi |
is there something speaking against? |
12:07 |
bernie |
cjb: I would expect those to be the same of sunjammer or slower |
12:07 |
silbe |
bernie: what was dfarnings estimate for us outgrowing sunjammer? would we (who?) have enough time for setting up clustered aslo without the "intermediate" solution? |
12:08 |
dogi |
has time to look if this machines are worth the time ... |
12:08 |
cjb |
bernie: you'd expect "Dual CPU 2.5 GHz AMD" to be slower than sunjammer? |
12:08 |
dogi |
even i have to travel to DC |
12:08 |
bernie |
silbe: he said we were doubling every 3 months |
12:09 |
|
silbe: I think it's more correct to say "we doubled in three months". growth seemed to be linear |
12:09 |
cjb |
dogi: not speaking against, it sounds good but it's a lot of machines. I think the only argument against would be that since we only need something like 1-3, it might be better to let them go to someone who's going to use them. |
12:09 |
bernie |
silbe: it is not easy to tell how much we'll grow in the future, but lately we've been seeing a *reduction* in traffic: |
12:09 |
|
http://sunjammer.sugarlabs.org[…]che_accesses.html |
12:10 |
cjb |
bernie: (oops, I got sunjammer and solarsail confused :) |
12:10 |
bernie |
I asked aa and rgs and they told me that schools are closed in uruguay and paraguay, which would explain the current trend |
12:10 |
silbe |
interesting, might be term-related |
12:10 |
bernie |
we can't easily predict how many more kids will use aslo over the next school year in latam |
12:11 |
|
cjb: sunjammer is a "Dual-Core AMD Opteron(tm) Processor 2216 HE" |
12:11 |
|
cjb: we also need to transition away from solarsail, btw |
12:11 |
cjb |
nod |
12:12 |
bernie |
cjb: it's too slow to run trac and the wiki, and the OS is aging with no good support in latest releases of ubuntu |
12:12 |
cjb |
yup |
12:12 |
dogi |
thats why i agree bernies idea to buy a new machine ala sunjammer, too |
12:12 |
bernie |
treehouse and housetree would be fast enough, but the hardware is consumer-grade... not sure we want to host really critical stuff there |
12:13 |
lfaraone |
Okay. Can we get agreement on the destinations of the 12 servers from WMF? |
12:13 |
bernie |
dogi: we have at least $2000 earmarked already, we'd have to get to $3000 |
12:13 |
|
lfaraone: right, let's get back on track |
12:13 |
|
lfaraone: are you in the loop with RIT? I guess not |
12:13 |
|
lfaraone: we should ask them if it's ok to host 3 1U machines there |
12:13 |
lfaraone |
bernie: no, I am not. |
12:14 |
silbe |
sounds like a bunch of new servers could help us expand and avoid buying a new machine from our limited funds (but we'd probably need some additional parts) in exchange for increased admin overhead |
12:14 |
lfaraone |
bernie: assuming they all work :) |
12:14 |
bernie |
lfaraone: since i'll be traveling for the next 25h, I'd prefer if you did the talking. |
12:14 |
|
#action bernie will forward email communication between us and rit to lfaraone |
12:14 |
|
I can't do action items? |
12:14 |
djbclark |
suggests a limited-access mailing list when dealing with sites that donate machines / services |
12:15 |
bernie |
dogi: you're chairing the meeting |
12:15 |
dogi |
no |
12:15 |
djbclark |
(seems to work well with lemote) |
12:15 |
lfaraone |
bernie: silbe is. |
12:15 |
silbe |
#action bernie will forward email communication between us and rit to lfaraone |
12:15 |
djbclark |
(that way if the person handling it normally has to stop, it's easy for another to see history and pick up) |
12:15 |
lfaraone |
silbe: +1, I think you can safely mark that #agreed |
12:16 |
|
djbclark: that sounds like a good use for RT :) |
12:16 |
bernie |
djbclark: systems lists.sugarlabs.org is kind of limited access. subscriptions require approval and archives are not public |
12:16 |
silbe |
#agreed |
12:16 |
djbclark |
throws up in the nearest available receptical |
12:16 |
bernie |
lfaraone, djbclark: I would not mind installing rt for sugarlabs. the problem is getting people to actually use it. |
12:17 |
dogi |
can create an rtVM on treehouse |
12:17 |
|
is there an interest |
12:17 |
djbclark |
lfaraone: RT is fail when working with externals; people add people to cc: and then tickets get duped. Roundup is a bit better. And perhaps newer RT, FSF has rather old version running. |
12:17 |
silbe |
isn't RT the system OLPC used for handling sending out the salvaged XOs? they managed to send each of their replies to a different email address of mine ;) |
12:18 |
dogi |
+1 silbe |
12:19 |
silbe |
so to get back on topic (i need to continue work on my thesis), what do we do about the servers? |
12:19 |
bernie |
djbclark, dogi, silbe: so let's explore other possibilities... who wants to do that? |
12:19 |
silbe |
do we want them? |
12:19 |
dogi |
yes |
12:19 |
bernie |
silbe: your email addressing scheme defeats me too btw :) |
12:20 |
silbe |
bernie: do what? |
12:20 |
|
bernie: :-P |
12:20 |
|
if defeats SPAM as well, though :) |
12:20 |
bernie |
silbe: can't you just get plenty of spam like anyone else? :-) |
12:20 |
dogi |
hopes to be able help dfarning in RIT |
12:22 |
cjb |
silbe: as long as WMF understands that we only plan on using ~2 of them, +1 on taking them from me |
12:22 |
bernie |
dogi: I'm not sure dfarning wants to be in the loop, but it would be great if you could convince him... |
12:22 |
lfaraone |
Do we want them in the places I've described earlier? (6 - {ACC, TWIX, Obscure}, 3 - RIT, 3 - Walter) |
12:22 |
|
cjb: SLDC will be sure to use all of the available servers :) |
12:23 |
silbe |
bernie: +1 on getting dfarning motivated again, it would help us a great deal |
12:23 |
djbclark |
bernie: I would suggest setting up a mailing list since that can be done with existing systems in a few minutes; if that ends up being annoying, can always set up a issue tracking system and redirect the list address there later. |
12:23 |
bernie |
silbe: I'll try to call him on the phone from the airport to say goodbye. |
12:24 |
silbe |
sorry to display my lack of geographic knowledge again - is any of the other places near DC? |
12:24 |
bernie |
djbclark: can't we just use systems@ for this? I'm trying to reduce the proliferation of dozens of deserted mailing lists, like at olpc |
12:25 |
cjb |
bernie: +1 |
12:25 |
bernie |
silbe: no, rit is in NY (a few hours drive west from boston) |
12:25 |
cjb |
bernie: btw, can I subscribe to systems@? is it just for people with root, or any infra team members? |
12:25 |
silbe |
then i'd say we should ship at least one box to DC |
12:25 |
bernie |
lfaraone: sounds good, pending authorization from RIT and TWIX |
12:26 |
djbclark |
bernie: unless the RIT people want to get a lot of unrelated email, that would seem inadvisable. Oh unless you mean like just be sure to always cc: systems and then set systems to allow posts by RIT people... that would work but have to be cognisant; also in that case, perhaps bounce existing emails to systems@? |
12:26 |
bernie |
lfaraone: can you check shipment costs? it may be quite high, since each machine probably weights ~20kg |
12:26 |
dogi |
lol |
12:27 |
silbe |
wow, what did they put in it? lead or something? |
12:27 |
cjb |
oh, I thought someone else was paying shipping :) |
12:27 |
|
yes, shipping will surely be hundreds of dollars |
12:27 |
lfaraone |
bernie: they estimated about 300usd for all of them, but they said that they couldcover it if it was a problem. |
12:27 |
djbclark |
bernie: I may be driving to DC area soon to visit a friend; this is RIT -> DC? |
12:27 |
lfaraone |
djbclark: right now they are in florida :) |
12:30 |
bernie |
walterbender: to explore the other path, we need to find out if we can afford $3000 on infrastructure |
12:30 |
lfaraone |
djbclark: i'm not sure if FL>DC is cheaper than FL>MA |
12:30 |
bernie |
djbclark: no, transworldix is in dc. rit is in rochester. |
12:33 |
lfaraone |
has to go to class... |
12:35 |
|
So I'll report back with the estimated shipping costs. |
12:36 |
dogi |
6 -> dc (lfarone), 3 -> rit (dfarning), 3 -> bos (walterbender) |
12:36 |
|
?? did i understand that right? |
12:37 |
|
+1 lfaraone |
12:37 |
bernie |
lfaraone: sent email to you and systems@ |
12:37 |
silbe |
dogi: i think so |
12:38 |
dogi |
thx |
12:38 |
|
:) |
12:38 |
|
cu |
12:38 |
bernie |
lfaraone: are you also going to write to twix to confirm they're still going to host us? they did not reply last time I wrote to them |
12:38 |
|
lfaraone: I hope they haven't changed their mind |
12:38 |
dogi |
s laptop is going out of power ... |
12:39 |
bernie |
let's wind down |
12:39 |
dogi |
#endmeeting? |
12:39 |
bernie |
sdziallas: ciao |
12:39 |
|
dogi: I guess meetbot did not like the ? |
12:39 |
dogi |
#endmeeting |
12:39 |
silbe |
#endmeeting |