July 5, 2012 
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | >
July 5, 2012 [00:00:13] <Licenser> try to fire it up?[00:00:19] <jeffpc> online :)[00:00:38] <Licenser> good :)[00:00:52] <Licenser> but it's odd that it didn't worked out of the box for you[00:01:14] <Licenser> can you do: svcs -x the three services wiggle, sniffle and snarl should be down[00:01:37] <jeffpc> wiggle & sniffle are in maintenance[00:01:46] <Licenser> snarl not - strage[00:01:53] <Licenser> well just clear them - they should now come up clean[00:02:38] <jeffpc> yep, all online[00:03:18] <Licenser> good snarl too?[00:03:21] <jeffpc> yep[00:03:41] <jeffpc> ooh, a web browser greeted me with a login page :)[00:03:41] <Licenser> just to be sure can you do /fifo/snarl/bin/snarl start[00:04:32] <Licenser> also you need to go out of the zone and do /opt/chunter/bin/chunter start[00:04:39] <Licenser> those two will not come up clean with redis down[00:05:15] <jeffpc> ok[00:05:16] <jeffpc> done[00:05:41] <Licenser> then you can login with admin and admin as password[00:06:21] <jeffpc> cool :)[00:06:30] <jeffpc> it lists the one zone (fifo)[00:08:13] <Licenser> that makes sense[00:08:16] <Licenser> next go to the admin page[00:08:24] <jeffpc> ok[00:08:29] <Licenser> there in the top left is an area with Ip info[00:08:35] <jeffpc> yep[00:08:51] <Licenser> you've to fill in 3 values Subnet/STart then Netmask then gateway[00:08:52] <jeffpc> set it?[00:08:53] <jeffpc> ok[00:08:55] <Licenser> not yet[00:09:00] <Licenser> one thing about the subnet start[00:09:34] <Licenser> if you take for example 192.168.0.100 with a subnet of /24 and a GW of 192.168.0.1 the following will happen[00:10:15] <Licenser> the first 99 addresses will be considered reserved and not handed out by fifo, the rest is fair game, you'll still see, after saving, the actual subnet which would be 192.168.0.0[00:10:39] <Licenser> now when you've understood the bad sentences you can go and set it ^^[00:10:44] <jeffpc> ok :)[00:10:58] <jeffpc> ok, set[00:10:58] <Licenser> because I know people get confused at that point - it's not ideal[00:11:05] <Licenser> now go back to the home page[00:11:17] <Licenser> and there is a + in the (empty) package sectiopn[00:11:20] <Licenser> there you can define packages[00:11:39] <jeffpc> ah, package == "some amount of service I want to 'sell'"[00:12:22] <jeffpc> ok, made one[00:12:37] <Licenser> use/sell/eat for brackfast[00:12:45] <Licenser> it's just a spec for the 'size' of a machine[00:13:01] <jeffpc> right[00:13:15] <Licenser> I went with the same names as joyent since I figured people are less confused like that :P[00:13:22] <jeffpc> yeah[00:13:24] <jeffpc> good idea[00:13:52] <Licenser> If I'd call them Specs, or Templates, or Pancakes someone using fifo going to SDC would be confused and visa versa - gain for noone there[00:14:04] <Licenser> now you're pretty much good to create your first machien[00:14:21] <jeffpc> there are no datasets listed[00:14:54] <Licenser> oh o.O darn[00:14:59] <Licenser> I think you have a point there[00:15:08] <jeffpc> imgadm lists:[00:15:09] <jeffpc> f9e4be48-9466-11e1-bc41-9f993f5dff36 smartos 2012-05-02 sdc:sdc:smartos64:1.6.3[00:15:12] <Licenser> I'm looking for the 'old' place[00:15:15] <Licenser> not the new one[00:16:00] * Licenser shakes his fist at konobi[00:17:25] <jeffpc> btw, does the fifo zone really need 2GB? :)[00:18:05] <Licenser> jeffpc you can propably shut it doen and give it 500m :P[00:18:11] <Licenser> or 1g[00:18:17] <Licenser> never measured it ^^[00:20:11] * jeffpc is going to give vmware more memory[00:22:01] <jeffpc> I gave it 1GB :)[00:22:13] * Licenser fixes[00:22:14] <jeffpc> it had 512MB before :)[00:22:18] <Licenser> heh[00:22:28] <Licenser> then you can give fifo 512[00:23:26] <jeffpc> FWIW, it looks like fifo didn't autostart[00:25:04] <jeffpc> oh, it did[00:25:10] <jeffpc> it's just slow at startup[00:25:20] <jeffpc> that's what I get for running it in vmware on a laptop[00:25:50] <Licenser> ^^[00:26:50] <Licenser> do you still have the fifofy script?[00:26:51] <jeffpc> how can I get the fixed version?[00:26:58] <jeffpc> thing so[00:27:11] <Licenser> okay then please do: /opt/chunter/bin/chunter stop[00:27:28] <Licenser> then ./fifofy.sh chunter <GZ IP> <fifo zone ip>[00:27:36] <Licenser> and /opt/chunter/bin/chunter start[00:28:11] <jeffpc> [root@00-0c-29-78-ec-df ~]# /opt/chunter/bin/chunter stop[00:28:11] <jeffpc> Node 'chunter at 10 dot 0.0.226' not responding to pings.[00:28:19] <jeffpc> that's the IP of the GZ[00:30:37] <Licenser> that is OK just do the next steps :)[00:30:45] <Licenser> it just means that chunter didn't came up correctly[00:30:55] <Licenser> I guess it had to do with the very slow startup[00:31:22] <jeffpc> ok[00:31:39] <Licenser> then you hopefully shoud see you datasets :)[00:31:45] <jeffpc> cool :)[00:32:27] <jeffpc> yep, I see the smartos64...[00:32:46] <Licenser> pew great :)[00:33:01] <Licenser> what I'm very proud of is the system page :)[00:33:05] <Licenser> you get live CPU usage stats there[00:33:16] <jeffpc> I saw :)[00:33:43] <Licenser> yay[00:33:52] <Licenser> it's fifo's crown jewel at the moment :D[00:34:13] <jeffpc> I find the "Home" page much more interesting :)[00:34:17] <jeffpc> because that's whan I want[00:34:25] <jeffpc> I'm going to make a new machine :)[00:34:49] <Licenser> heh[00:35:07] <Licenser> if it were a real host you'd even get a VNC button that opens a noVNC clinet for it :)[00:35:37] <jeffpc> oooh :)[00:36:30] <jeffpc> svc.configd is chewing up a bunch of cpu[00:37:29] <Licenser> that isn't mine :P[00:37:35] <jeffpc> I know[00:38:04] <jeffpc> hrm, sometimes wiggle displays an empty-ish page[00:38:11] <jeffpc> like system with no content[00:38:18] <jeffpc> or home with only the table but no entries[00:39:58] <Licenser> hmm odd o.O[00:40:20] <jeffpc> I don't know if that's because the vm is pretty slow or what[00:40:39] <Licenser> might be that the data just isn't loaded then[00:41:32] <Licenser> but you'd have to look into the error console of the browser if you can[00:42:14] <jeffpc> some css issues acording to firefox[00:42:51] <jeffpc> [17:42:34.527] The connection to ws://10.0.0.18/events was interrupted while the page was loading. @ http://10.0.0.18/js/application.js:406[00:44:21] <jeffpc> hrm[00:44:37] <jeffpc> it actually looks like wiggle is not sending back a list of machines & datasets[00:44:43] <jeffpc> the one package does show up[00:45:00] <Licenser> o.O[00:45:01] <Licenser> odd[00:45:38] <Licenser> can you do a ./fifofy.sh collect there is a .tar.bz coming out of that if you mail it to me I'll look into it tomorrow (it's to late today I need to get to bed)[00:45:58] <jeffpc> sure[00:46:01] <jeffpc> where do I send it?[00:47:34] <Licenser> I send you my email in a query :)[00:47:58] <Licenser> one should net tempt the spambots![00:48:13] <jeffpc> :)[00:48:28] <Licenser> I'll head to bed now sorry we couldn't get it working fully :/[00:48:31] <jeffpc> it's 166KB :)[00:48:34] <Licenser> I'll try to dig it out[00:48:37] <Licenser> yea lots of logs[00:48:37] <jeffpc> no problem[00:48:44] <jeffpc> thanks for the help[00:48:53] <Licenser> for the development purpose I've a rather extensive logging[00:49:17] <Licenser> You're more then welcome mate :) thanks for trying fifo[00:49:22] <Licenser> so wiht this I'm off to bed![00:49:23] <Licenser> nigt people[00:49:25] <jeffpc> night[00:55:35] *** Orpheus has joined #smartos[00:55:51] <Orpheus> hey is there plans of making smartos work with amd?[00:56:21] <bdha> Orpheus: There is a fork. It works, but it doesn't get much love.[00:56:32] <Orpheus> why not?[00:56:39] <Orpheus> it works but doesnt work well?[00:56:46] <bdha> Oh, this is #smartos, not #illumos.[00:56:50] <bdha> Sorry.[00:56:56] <Orpheus> lol[00:57:05] <bdha> There is a fork of the illumos code that works on AMD. But it doesn't get a lot of work, because people have other priorities.[00:57:20] <bdha> Joyent doesn't want to do the work, because their entire fleet is Intel.[01:00:33] <jeffpc> apparently, the code works for some people[01:00:44] <jeffpc> which isn't good enough to integrate[01:01:10] <bdha> Sure. And it would require maintenance.[01:01:17] <bdha> No one has time to take ownership of it.[01:01:46] <jeffpc> I might try to hack at it[01:01:55] <bdha> Cool.[01:01:57] <jeffpc> it depends if I can dig up another amd box[01:02:13] <jeffpc> the one I'd want to use smartos on is already in production[01:07:10] <richlowe> the reason it doesn't get much love is largely that the hardware people can reach works.[01:07:58] <richlowe> find a completely remote managable AMD that doesn't work, and some motivation, and I can look more.[01:08:12] <richlowe> "completely" would involve remote console, reset, and NMI[01:10:50] *** jamesd__ has quit IRC[01:14:45] *** denizr has joined #smartos[01:18:01] <rmustacc> It only working for some people and you know, no one saying that they think it's done and stable and want it to be integrated.[01:25:27] *** denizr has quit IRC[01:32:16] <nahamu> richlowe: are there any notes somewhere describing which systems it does work on?[01:35:04] <richlowe> nahamu: it's not something that fits any obvious pattern.[01:35:21] <richlowe> nahamu: I used it on an x2200-m2, which is basically the crappiest oldest CPU possible, and it was happy. LeftWing was on something far newer[01:35:34] <richlowe> in the gap between them, bad things happen, perhaps depending on system configuration, or CPU model, or ...[01:35:43] <richlowe> rmustacc: right, it certainly doesn't work well enough.[01:36:03] <richlowe> rmustacc: but it never will, unless someone with a broken system debugs it, or has a broken system featureful enough to let someone else[01:37:22] <richlowe> NPT is more likely to work, and where it didn't work, didn't crash the host[01:37:31] <richlowe> without NPT there's a known but undiagnosed bug that _does_ crash the host.[01:37:43] <rmustacc> I should probably give it a shot on my local machine at some point.[01:37:57] <rmustacc> But that's a problem of hard drive shuffling, etc.[01:39:08] <richlowe> rmustacc: if you have NPT, just booting Josh's iso and trying to boot a VM image perhaps over NFS would be sufficient[01:39:21] <richlowe> rmustacc: if it fails like it did on e^ipi, it'll refuse the (probably) very first entry[01:39:24] <Orpheus> hey im at the smartos config and i dont want to use dhcp. i put the ip address already. the netmask is the one my router lists right[01:39:27] <Orpheus> ?[01:39:41] <Orpheus> i really should brush up on networking :/[01:41:28] <jeffpc> richlowe: what's this ISO you speak of?[01:42:38] <jeffpc> Orpheus: that's the ip looking thing with many 255's[01:42:55] <Orpheus> yep[01:43:01] <Orpheus> oh many?[01:43:20] <richlowe> jeffpc: Josh built a smartos ISO with the amd-able/pre-ept-able kvm[01:43:22] <jeffpc> well, it depends on your setup[01:43:39] <Orpheus> it shows just 255.0.0.0[01:43:42] <jeffpc> richlowe: ooh, can I get that?[01:43:47] <jeffpc> Orpheus: ok, that's it[01:43:54] <Orpheus> ah alright :)[01:43:56] <rmustacc> He should probably build a new one.[01:43:57] <jeffpc> Orpheus: you just happen to have a big subnet[01:44:03] <richlowe> LeftWing: I shamefully don't know where you put the smartos iso[01:44:22] <LeftWing> It's pretty old now...[01:47:00] *** deirdres has quit IRC[01:48:44] <Orpheus> default gateway ip is just the routers ip address right?[01:49:43] <jeffpc> yep[01:57:30] <Orpheus> idk why but every illumos based os i try in qemu overuses my ram even though i had it restricted to like 1 gig.[01:57:39] <Orpheus> qemu/kvm[01:57:50] <rmustacc> As in the qemu process itself is using more then you expect?[01:57:59] <rmustacc> Or the overal system free memory is less then you expect?[01:58:46] <Orpheus> it is using way more then i expect[01:58:50] <Orpheus> like all of it[01:59:06] <rmustacc> The process or the system?[01:59:26] <Orpheus> the process[01:59:46] <jeffpc> kvm needs a bit for itself[01:59:49] <rmustacc> What's the size and rss on it?[02:00:24] <Orpheus> 20 gigs of space and rss?[02:00:52] <rmustacc> That seems busted.[02:01:29] <Orpheus> i cant even boot them. as soon as i do my ram usage just sky rockets[02:01:53] <jeffpc> Orpheus: KVM reserves all the RAM up front[02:01:56] <Orpheus> linux, bsd and windows OSs work fine[02:02:30] <jeffpc> (I may be misunderstanding what's going on)[02:03:26] <Orpheus> it shows just 255.itboaram even though i limited it to 1gig[02:03:34] <Orpheus> fail[02:03:53] <Orpheus> jeffpc: im trying to use illumos based OSes in qemu/kvm but as soon as i boot up the vms they take all the ram even though i limted to 1gig[02:04:08] <rmustacc> How're you getting that data?[02:04:14] <Orpheus> conky[02:04:15] <rmustacc> And how are you constructing the vm.spec you passed to vmadm?[02:04:15] <jeffpc> Orpheus: ouch[02:04:25] <rmustacc> And how is conky grabbing it?[02:04:31] <Orpheus> and the fact my computer grinds to a halt[02:04:42] <Orpheus> membar ;)[02:05:02] <rmustacc> It's running membar?[02:05:12] <Orpheus> you know what conky is right?[02:05:16] <rmustacc> Yes.[02:05:23] <rmustacc> Though I haven't used it for years.[02:05:39] <rmustacc> Sorry, just trying to understand some of the data and correlate it.[02:05:52] *** sebasp has quit IRC[02:06:25] <Orpheus> Mem: $mem/ $memmax[02:06:33] <Orpheus> ^ that is actually how i know it uses all my ram[02:06:46] <Orpheus> it lists how much ram i have and how much is being used.[02:08:58] <rmustacc> So if you're on the host[02:09:13] <rmustacc> Can you 'echo ::memstat | mdb -k'[02:10:22] <Orpheus> as soon as it boots up and gets past the menu my computer grinds to a halt and i cant do anything except for try to desperatly close the vm window[02:11:02] <Orpheus> well im off to eat 4th of july dinner. im sorry i goto go while you are helping me :/[02:11:06] <Orpheus> thanks you so far[02:11:24] <rmustacc> No problem.[02:11:38] <rmustacc> Come by tomorrow during office hours and someone can help you out further figure out what's happening.[02:21:43] <nahamu> LeftWing: is the iso basically just smartos-live with a different kvm repository?[02:23:14] <LeftWing> I unpacked an actual SmartOS release ISO and replaced the kvm module and /etc/issue[02:23:41] <LeftWing> (and then repacked it)[02:25:01] <nahamu> cool[02:36:19] <killfill> how do you guys normally send nginx or apache logs to a log server?..[03:09:04] *** dis has quit IRC[03:10:51] *** dis has joined #smartos[03:11:06] <Orpheus> huh odd smartos works fine in qemu now after a second boot. lol i still get errors. hehe works better then the one i had on my desktop ^_^[03:11:20] <Orpheus> i think im going to wait until i get a rack server.[03:11:28] <Orpheus> just keep openindiana on my desktop.[03:30:33] *** dis has quit IRC[03:32:34] *** dis has joined #smartos[03:40:15] *** denizr has joined #smartos[03:58:07] *** deirdres has joined #smartos[04:07:32] *** bixu has joined #smartos[04:10:30] <LeftWing> killfill: do you mean in real time? I've never had a requirement to do anything more complicated than the usual log rotation coupled with a "scp this to a holding area".[04:21:08] <nahamu> I've never used it, but I've heard good things about logstash[04:21:11] *** bixu_ has joined #smartos[04:22:18] *** bixu has quit IRC[04:22:51] <konobi> killfill: there's also logadm to look at[04:23:16] <richlowe> logadm would presumably be the motor behind what LeftWing said[04:32:12] <konobi> yeah... though with smarots style stuff, it's generally not registered with logadm[04:32:16] <konobi> =0/[04:36:40] <killfill> ah.. :)[04:36:58] <jeffpc> can't you write a five-liner in node that logs into gmail for you and sends you an email with each log entry? :P[04:37:17] <killfill> heh.. no thanks.. :P[04:38:41] <killfill> As im trying to migrate a platform to smartos.. i see things 'previously configured'.. like having apache log to file on a remote NFS[04:38:55] <killfill> dont think its a good idea..[04:39:08] <jeffpc> heh[04:39:23] <jeffpc> I don't know if there's an easy way to do remote logging[04:40:01] <jeffpc> in linux I'd probably set up syslogng to send everything to a logging server[04:40:21] <richlowe> With apache, etc, you can (iirc) have it log to syslog[04:40:30] <richlowe> and at least limit yourself to only one problem[04:41:14] <killfill> there has to be a plugin that send logs to a rabbitmq or something..[04:41:29] <killfill> anyway.. thanks :)[04:44:04] <konobi> well, syslog can do server stuff[04:50:44] *** denizr has quit IRC[04:53:25] <LeftWing> logadm is the work of the devil. install logrotate from pkgin. :P[04:54:35] <richlowe> <richlowe> I actually have code in logadm![04:54:42] <richlowe> So thanks, man.[04:54:56] <richlowe> (inconsequential code, but still[04:57:40] <jeffpc> heh[04:57:46] <jeffpc> I have code in man(1)![04:57:48] <jeffpc> :P[04:58:18] *** dis has quit IRC[05:00:04] *** dis has joined #smartos[05:01:13] <LeftWing> richlowe: I'm sure your bits are fine.[05:01:38] <LeftWing> I object mostly to the insane notion that configuration *and* state are kept in the *same* file, updated by both the user and the software and hopefully not at the same time.[05:01:50] <LeftWing> Also, if your /etc is full we'll helpfully truncate the file for you.[05:01:51] <LeftWing> Christ.[05:02:20] <LeftWing> I would be a thousand times happier if config was in /etc/logadm.conf and state was in /var/lib/logadm.state or something.[06:24:36] *** jamesd__ has joined #smartos[06:38:17] *** Orpheus has quit IRC[06:38:47] *** Orpheus has joined #smartos[06:53:40] *** darjeeling has quit IRC[07:13:39] <richlowe> file a bug[07:16:16] <richlowe> but I'm not seeing surprising stuff in there that's not config[07:22:17] <LeftWing> I'll put it on my list of things to look at. It's been some time since I've actually relied on it, and the brokenness would have been in the S10 version.[07:23:02] <LeftWing> richlowe: the -P flag is the most egregious violation of my expectations, though.[07:28:25] <richlowe> Yeah, I saw it in the manual page, I guess the default config doesn't necessitate it[07:28:27] <richlowe> and/or it's broken[07:28:47] <richlowe> LeftWing: and "your list" should consist of filed bugs, dammit.[07:31:44] <LeftWing> I file sometimes :P[07:39:36] <LeftWing> (and I think the -P is in response to a request for -p style periodic rotation)[07:39:43] *** darjeeling has joined #smartos[07:39:58] <MerlinDMC> morning[07:40:05] <LeftWing> hi[07:55:46] *** marsell has quit IRC[08:03:22] *** Licenser has quit IRC[08:06:02] *** marsell has joined #smartos[08:07:03] *** marsell has quit IRC[08:39:30] *** marsell has joined #smartos[09:01:23] *** Guest73363 has joined #smartos[09:04:42] *** Licenser has joined #smartos[09:15:58] *** Guest73363 is now known as max`[09:16:37] *** max` is now known as mrb[09:24:39] <Licenser> morning[09:26:33] *** bens1 has joined #smartos[09:35:17] <MerlinDMC> morning[09:37:01] <akole> good morning[09:40:11] *** mrb has quit IRC[09:44:29] *** AikoM has joined #smartos[09:46:45] <Licenser> MerlinDMC you were talking about not liking the gauges what is better?[09:54:42] *** jamesd__ has quit IRC[10:11:59] <MerlinDMC> Licenser, if you display only one host it's ok ... if it's a more than one host thing the blog entry of whoeverthatwas looked nice[10:12:37] <MerlinDMC> btw ... read the pull request -.-[10:15:49] <Licenser> oh you've a pull request?[10:16:07] <Licenser> which project?[10:16:48] *** _alram has joined #smartos[10:18:41] <Licenser> Iah found it[10:21:01] <MerlinDMC> i don't think that replacement will be enough for the change ... but i also have nothing installed here to compile that one and test it[10:23:16] <Licenser> yea vmwebadm is pretty much dieying anyway but I merged it non the less :)[10:23:28] <Licenser> the new SmartOS broke lots of stuff or me :([10:23:39] <Licenser> imgadm/dsadm then they removed node from path :/[11:24:22] <jkj> anybody running with jumbo frames and kvm?[11:24:51] <jkj> for example: ping -sD <some-ip> 5000[11:25:41] <jkj> there seems to be problems getting over 4096 byte packets over kvm interfaces[11:36:55] <MerlinDMC> jkj, could i try to ping the kvm from the GZ and get those broken results?[11:39:01] <jkj> 22:21:38.741060 IP truncated-ip - 1 bytes missing! (tos 0x0, ttl 64, id 0, offset 0, flags [DF], proto ICMP (1), length 4083) 10.0.0.125 > 10.0.0.32: ICMP echo request, id 3425, seq 17, length 4063[11:39:36] <jkj> this is the other way round. from kvm to gz.. snooped from global[11:40:55] <MerlinDMC> and you have enabled jumbo frames for the guest and the GZ?[11:55:42] *** v12 has joined #smartos[11:59:54] <jkj> MerlinDMC: yes. the mtus are set to 9000 from both sides. over 1500 byte packets pass.. but over 4082 byte packets get truncated[12:00:16] <jkj> there is 14 byte slack to page size... maybe some header is 14 bytes[12:08:52] <arekinath> arduino usually run off internal rc oscillators[12:59:53] *** _alram has quit IRC[13:29:00] *** _alram has joined #smartos[13:42:27] *** _alram has quit IRC[13:53:45] *** _alram has joined #smartos[14:13:07] *** _alram has quit IRC[14:50:09] <jeffpc> Licenser: one thought about the IP assignments... one should really need to enter a set of IP ranges[14:51:55] <jeffpc> Licenser: further down the line, it'd be nice to have VLAN support as well[14:53:11] <Licenser> jeffpc I totally agree, it makes more sense also allow different kind of netwoeks (external admin etc)[14:55:50] *** ira has joined #smartos[14:58:04] <jeffpc> Licenser: did you get a chance to figure out what's the deal with my problematic fifo?[14:59:04] *** darjeeling has quit IRC[15:03:12] <Licenser> not yet, still sitting in the basement trying to get a system to run - dayjob and all that icky stuff ::P[15:04:40] <jeffpc> ah[15:04:50] <jeffpc> let me know if you need more info[15:14:10] *** SaskaloonAFK has joined #smartos[15:15:34] *** SaskaloonAFK has quit IRC[15:17:47] <Licenser> jeffpc will do no worries :)[15:25:45] *** Saskaloon has joined #smartos[15:26:54] *** denizr has joined #smartos[15:28:38] <Licenser> jeffpc also if you have any ideas/suggestions/wishes feel free to put them in the github tracker (https://github.com/project-fifo/fifo) it helps me a lot to keep suggestions sorted :)[15:29:49] <jeffpc> will do[15:30:35] <nahamu> Licenser: were you working on a replacement for vmwebadm that talks to chunter?[15:30:51] <Licenser> nahamu yap[15:31:36] <nahamu> what's that called?[15:33:08] <Licenser> nahamu no name yet ^^[15:33:12] <Licenser> but not vmwebadm that is for sure[15:33:22] <Licenser> perhaps pounce, not sure yet[15:46:10] *** CarlosC has joined #smartos[15:47:10] *** jamesd has joined #smartos[15:57:28] <MerlinDMC> where do i get 32bit virtio drivers for windows (xp, 2003, 2008)? smartos-vmtools only include 64bit win7 and up i think[15:59:14] *** darjeeling has joined #smartos[15:59:42] <rmustacc> Probably Fedora/Redhat.[15:59:54] <rmustacc> If they exist. Not sure if they do or not.[16:01:04] <rmustacc> We didn't implement the virtio drivers.[16:04:40] <nahamu> http://alt.fedoraproject.org/pub/alt/virtio-win/latest/images/bin/virtio-win-0.1-22.iso perhaps[16:04:52] <nahamu> (found via http://www.linux-kvm.org/page/WindowsGuestDrivers/Download_Drivers )[16:05:11] <nahamu> oh, that says those are for 64bit... oops.[16:05:38] <nahamu> hmmm[16:07:56] <nahamu> This page might have them... I didn't test the links: http://blog.famzah.net/2010/01/09/kvm-qemu-virtio-storage-and-network-drivers-for-32-bit64-bit-windows-7-windows-vista-windows-xp-and-windows-2000/[16:09:40] *** neophenix has joined #smartos[16:15:46] <MerlinDMC> nahamu, those include x86 also ... but i didn't know if there would be a better source[16:16:16] <MerlinDMC> rmustacc, would it hurt to make them also available in smartos-vmtools?[16:17:32] <rmustacc> If we can get them directly from upstream, it might make sense.[16:17:39] <rmustacc> I don't remember who does the actual signing in these cases.[16:17:47] <rmustacc> The ones on that blog entry I wouldn't particularly trust.[16:18:30] <MerlinDMC> is trevoro available to get poked on that one? ;)[16:18:42] *** enmand has joined #smartos[16:18:56] <rmustacc> Nope[16:19:11] <rmustacc> He's on leave for a while.[16:20:27] <MerlinDMC> hm ... I'll try to play around with those from redhat first[16:20:36] <nahamu> you could be sneaky and go ask the fedora people where to get drivers.[16:20:52] <MerlinDMC> but i'll also take a note on that for later[16:39:16] *** elijah-mbp has joined #smartos[16:41:13] <ismell> is there a net virtio driver yet ?[16:41:26] <MerlinDMC> ismell, yop[16:41:33] <ismell> ohh, when did that get included ?[16:41:52] <MerlinDMC> i don't know ... did download the redhat iso thing today[16:42:19] <MerlinDMC> the smartos-vmtools also have a netkvm[16:45:01] <ismell> is the vmware tarball in the download made for running as a vm ?[16:45:17] <MerlinDMC> ismell, yes[16:45:48] <ismell> ohhh[16:46:02] <ismell> might have to try that out on my home box[16:46:09] <MerlinDMC> inside vmware ... works[16:46:33] <MerlinDMC> is nothing else than a preconfigured vmware file with an included cd iso[16:46:39] <ismell> I'm running VirtualBox right now :([16:46:48] <MerlinDMC> but nice to have and test smth quick[16:47:29] * MerlinDMC driving home now[16:57:49] *** neophenix_ has joined #smartos[16:59:52] *** neophenix has quit IRC[16:59:52] *** neophenix_ is now known as neophenix[17:03:37] *** jdavis has joined #smartos[17:12:23] *** ryancnelson has joined #smartos[17:24:44] *** ryancnelson has quit IRC[17:26:29] *** scanf has quit IRC[17:26:52] <nahamu> are there any signs in the internal Joyent repos that ipv6 support might be landing soon?[17:27:20] <rmustacc> In vmadm?[17:27:26] <nahamu> yeah[17:27:55] <rmustacc> Not in the immediate short-term.[17:28:13] <nahamu> well, vmadm and the brands[17:28:14] <nahamu> okay[17:29:49] *** ryancnelson has joined #smartos[17:38:59] *** AikoM has quit IRC[17:43:48] *** Licenser has quit IRC[17:50:21] *** Licenser has joined #smartos[18:01:18] *** max has joined #smartos[18:01:42] *** max is now known as Guest49589[18:02:02] *** Guest49589 is now known as mrb[18:02:56] *** knweiss has joined #smartos[18:03:22] *** pijewski has joined #smartos[18:12:20] *** bens1 has quit IRC[18:37:17] *** ryancnelson has quit IRC[18:46:21] *** sebasp has joined #smartos[18:50:04] *** JoshWilsdon has joined #smartos[18:52:54] *** jdavis has quit IRC[18:55:04] *** elijah-mbp has quit IRC[18:59:48] *** Prudhvi has quit IRC[19:00:29] <rmustacc> Welcome to office hours.[19:02:41] <nahamu> e^ipi: were you able to play with the build_usb example?[19:03:03] *** rahim_ has joined #smartos[19:04:44] <e^ipi> not yet, no[19:08:31] <d[^_^]b> am i the only one who feels like SmartOS KVM is more responsive and has better scheduling capacity than Linux KVM?[19:09:11] <d[^_^]b> i don't understand why though[19:09:16] <rmustacc> Unfortunately there's no way to verify it.[19:09:22] <rmustacc> Well, we have an entirely different schedular.[19:09:39] <rmustacc> Well, an entirely different kernel.[19:10:16] <d[^_^]b> i ran linux kvm and smartos with same workload on the same hardware (DL380G6, 12 cores, 72GB RAM, hardware RAID6 with only ZFS on top of it) and SmartOS always has lower host zone load, less latency inside guests[19:10:52] <rmustacc> Great.[19:11:01] <d[^_^]b> the box also runs cooler.[19:12:16] <killfill> that would be a greate slogan.. "SmartOS runs cooler".. :P[19:14:02] <d[^_^]b> it's a bummer that i can't use it on our older servers, poweredge 2950 but luckily those are out the door and we're getting G8's instead[19:24:08] *** mrb has quit IRC[19:25:30] <sebasp> Did anyone test Smartos on Cisco UCS?[19:27:56] <rmustacc> I know some folks have run it on some Cisco boxes before.[19:28:08] <rmustacc> sebasp: Do you have a BOM for what you want to run on?[19:29:06] <sebasp> Thinking of getting 2 x C200 with 4 drives to run SmartDC[19:29:31] <rmustacc> You'll want to double check the disk controllers and nics against the HCL.[19:30:14] <sebasp> I currently run on X2270 for Sun…trying to find something around the same type of machine[19:30:49] <sebasp> rmustacc: thanks, will do check against the HCL.[19:31:09] <nahamu> random question: any idea when the next sync-up with illumos will be?[19:31:19] <rmustacc> If you're uncertain if something is on there, etc. you can ping us with the device name and ideally the pci id, and it'd be pretty straightforward.[19:31:28] <sebasp> ok...[19:31:31] <rmustacc> nahamu: In the next week or two so I'd imagine.[19:31:48] <nahamu> cool[19:33:22] <CarlosC> another random question: what are some of the "killer" new features Joyent is bringing to SmartOS?[19:33:54] <sebasp> CarlosC: Just smartos … ain't enough :)[19:34:06] <CarlosC> that works ;)[19:34:40] <Licenser> I wonder Is there a reason that that disks for KVM's are done as /zones/<uuid>-diskX instead of /zones/<uuid>/diskX ?[19:35:06] *** dap has joined #smartos[19:35:23] <rmustacc> CarlosC: I'm working on some improvements to the userland memory allocator.[19:35:24] <CarlosC> with the recent DTrace changes to allow inter-zone usage, what is the vision/goal to bring more visibility to KVM?[19:35:47] <richlowe> Licenser: there was some code in vmadm (in the past) that forced them to be only one level deep, also[19:35:54] <richlowe> Licenser: bit bdha at some point, I think it got fixed.[19:36:02] <richlowe> rmustacc: how's that going, by the way?[19:36:06] <Licenser> ah I see :)[19:36:29] <rmustacc> CarlosC: Well, some day we'd love to do something like vfbt.[19:36:39] <rmustacc> It's just a matter of time, priorities, and getting angry enough.[19:36:52] <rmustacc> richlowe: Prototypes are working. So it's a lot of benchmarking, etc. now.[19:36:55] <CarlosC> rmustacc it's not that VM2 that was on the list a while ago is it?[19:37:23] <rmustacc> That's enitrely different.[19:37:32] <rmustacc> VM2 is the OS Virtual memory subsystem.[19:37:37] <CarlosC> ah...[19:37:40] <rmustacc> This is just libumem enhancements.[19:37:48] <CarlosC> nice...[19:37:52] <rmustacc> For good 'ol malloc(3C) and free(C)[19:38:23] <CarlosC> what are improvements? different strategies, etc..?[19:38:59] <rmustacc> Essentially per-thread caching of recently used small allocations so you can get them in a lock-free manner.[19:39:16] <killfill> on my smartmashine, node's os.cpus() reports 8 cpus, wich has the compute node. is that correct?[19:39:20] <rmustacc> But he devil is in the details, so you can observe them all when it's done.[19:39:33] <rmustacc> killfill: Run psrinfo -vp.[19:39:39] <rmustacc> That'll show you all the CPUs you have.[19:39:45] <rmustacc> os.cpus() should report all of them.[19:40:06] <killfill> hm.. i get a nice perl error.. :S[19:40:27] <rmustacc> Hmm, really?[19:40:32] <rmustacc> What's it saying?[19:40:39] <killfill> http://pastebin.com/JyhnqDzQ[19:41:11] <rmustacc> Okay, that's pretty weird.[19:41:39] <rmustacc> Are you running SmartOS?[19:41:42] <killfill> i got the latest percona dataset i found, that is 1.5.1[19:41:52] <wesolows> what platform are you on?[19:42:08] <killfill> oh its a plain joyent sdc install[19:42:24] <killfill> 5.11 joyent_20120614T001014Z i86pc i386 i86pc Solaris[19:42:25] <rmustacc> So you installed SDC 6.5?[19:42:29] <killfill> yup[19:42:34] <killfill> its a smartmashine[19:42:51] <rmustacc> Just to rule out something stupid, if you run psrinfo -vp from the global zone, what happens?[19:43:32] <killfill> oh, the same[19:43:53] <rmustacc> Okay, hmm.[19:44:05] <killfill> linke "use 5.006_001;"[19:44:09] <killfill> line[19:44:38] <richlowe> what's your locale?[19:44:52] <richlowe> I've seen this before, but don't quite remember how, so that might be a silly question[19:45:01] <richlowe> but try forcing LANG and LC_ALL to "C"[19:45:07] <killfill> oh its the locale...[19:45:10] <rmustacc> Oh.[19:45:18] <richlowe> iirc, we've had problems before where perl got botched to parse version numbers using the locale fraction separator[19:45:24] <richlowe> so if you were german, it'd want '5,6'[19:45:29] <killfill> the instalation somehow put the 'right' locale based on my location or something[19:45:48] <rmustacc> Sometimes ssh forwards that on.[19:45:48] <killfill> im es_CL.UTF-8 :P[19:45:57] <killfill> aah that has to be[19:46:53] <killfill> hm.. i see this: http://pastebin.com/KPZ8CSzP[19:47:18] <killfill> whats does it mean?.. the vm is using all the cpu's?[19:48:02] <rmustacc> No.[19:48:26] <rmustacc> So, you have to remember that the smartmachines aren't doing hardware virtualization like traditional VMware/KVM, etc.[19:48:35] <richlowe> rmustacc: might want to investigate your perl build, if the locale is tripping like that. It can't be how things _should_ work.[19:48:50] <rmustacc> richlowe: Yeah, I'm sure there's something weird going on there.[19:49:19] <rmustacc> killfill: So, in this case, the smart machine sees all of the hosts CPU resources, but it's limited in what it can use based on its shares and caps.[19:49:39] *** jdavis has joined #smartos[19:50:42] <killfill> ok, and that should be configurable[19:51:09] <rmustacc> It is.[19:51:15] <rmustacc> In SDC, you want to look at packages.[19:52:30] *** enmand has quit IRC[19:53:51] <killfill> Ah there it is. i was just looking at rezise option of the vm..[19:55:23] *** BartL has joined #smartos[19:57:09] *** deirdres has quit IRC[19:58:27] <BartL> anyone here know where I can file a 'bug' in vmadm? (mostly an annoying typo)[19:58:56] *** enmand has joined #smartos[19:59:29] <MerlinDMC> BartL, github.com/joyent/smartos-live/issues[20:00:07] <BartL> thanks[20:03:14] *** ryancnelson has joined #smartos[20:06:12] <akole> possible stupid question, but is there any known issues when using aggr as the external interface for VMs?[20:08:53] <rmustacc> For KVM VMs?[20:09:28] <akole> yes[20:10:47] <rmustacc> I don't know of anyone trying to use it.[20:10:50] <ryancnelson> if you consider "doubling or quadrupling your network complexity, and reducing the benefits of a cloud architecture" to be "issues", then yeah.[20:11:24] <ryancnelson> you need aggr for the bandwidth, or for redundancy?[20:11:30] <akole> redundancy[20:11:50] <ryancnelson> if it's redundancy, if a nic fails, move your nictag, and reboot[20:11:57] <ryancnelson> that's "the cloudy way"[20:12:13] <akole> ...doh... didn't even think of that >_<[20:12:50] <rmustacc> aggrs should be able to work.[20:12:53] <rmustacc> If not, it's a bug.[20:13:20] <ryancnelson> right… the tech is in there, it's just way more complicated to orchestrate[20:13:32] <MerlinDMC> ryancnelson, you always reboot in case of errors right? ^^ ... i liked that video describing why smartos is a liveimage[20:14:02] <ryancnelson> which? bryan cantrill's video? or the brief bit in my install vid?[20:14:16] <ryancnelson> if you blow up a nic, yeah, you pretty much need to reboot.[20:14:53] <MerlinDMC> that one with bryan asking questions[20:14:55] <ryancnelson> … when you boot a vm or a zone, the nics are created on the fly. so if you've redefined your nic tag, the place where it decides what physical nic to build your vnic on changes, post reboot[20:16:00] <ryancnelson> yeah, that was done early on when we had to explain it a lot. it's not a new idea, in the world of hypervisor deployment, but people who love solaris don't tend to think of it as a thin layer on top… it's too full-featured :)[20:20:59] *** deirdres has joined #smartos[20:37:27] <Tekni> ryancnelson: we must unlearn what we have learned :)[20:37:53] <ryancnelson> tell. me. about. it.[20:38:08] <ryancnelson> story of my lfe :)[20:56:53] *** sebasp has quit IRC[21:00:02] *** trentm has joined #smartos[21:08:37] *** Licenser has quit IRC[21:24:24] *** knweiss has quit IRC[21:36:18] *** BartL has quit IRC[21:37:15] *** ryancnelson has quit IRC[21:39:27] *** sebasp has joined #smartos[21:41:59] <killfill> is it common to test mysql preformance with sysbench?...[21:42:47] <e^ipi> aren't there specially tailored mysql benchmarks?[21:43:51] <killfill> im fighting with a centos@xen vm... sysbench tells its faster than percona(smartos) at sdc dot . :S[21:45:29] <killfill> e^ipi, wich ones?..[21:46:19] <e^ipi> mysqlbench ?[21:47:36] <killfill> hm..[21:58:12] *** neophenix has quit IRC[21:59:17] <jeffpc> I'd expect ext*/xfs on linux to outperform zfs, just because they don't COW[21:59:31] <jeffpc> killfill: define "faster"[22:00:10] <killfill> well, i did NFS mount from the same NAS on both mashines... the databases are located in the NAS mount[22:00:15] <jeffpc> ah[22:01:00] *** Licenser has joined #smartos[22:01:13] <killfill> the centos is a mysql 5.1.61 tho..[22:01:58] <killfill> mysql@smartos is 5.1.58[22:02:35] <killfill> jeffpc: i mean faster by sysbench reporting it does more tps and red/writes[22:03:02] <killfill> Anyway.. i know i need to make a better bench.. :P[22:03:06] <jeffpc> killfill: how much faster? 1% 2%...[22:03:25] <jeffpc> killfill: no, there are too many benchmarks as it is[22:03:34] <rmustacc> killfill: I would probably by starting the benchmark against a database on the local filesystem and make sure it's either cached or not, so you can rule out a lot of stuff.[22:03:39] <jeffpc> killfill: you need to use representative workload :)[22:03:49] <rmustacc> Otherwise right now you have to diagnose and verify your network, nfs client, nfs server, etc.[22:04:36] <killfill> smartos gives 80tps, and centos is giving 100tps[22:04:50] <jeffpc> ok, significant enough[22:04:58] <rmustacc> The things you want to start looking at are where you're hitting the bottlenecks.[22:05:05] <rmustacc> Is it on the network, is it CPU, is it memory, etc.[22:05:21] <rmustacc> Are you using similar tuning for mysql on each, etc.[22:05:29] <killfill> rmustacc, i tested on a local fs too.. the result is not very different[22:06:05] <rmustacc> Then I'd focus on the local fs one, as that's a lot easier to look at and understand.[22:06:13] <rmustacc> Less variables.[22:06:34] *** neophenix has joined #smartos[22:06:46] <killfill> yup, they have exactly the same my.conf, except that the percona needed to have these two options, if not it didnt started : default-storage-engine = innodb innodb_log_file_size = 400M (i checkout that the test databases are both innodb)[22:07:10] <killfill> right[22:09:34] <killfill> how do i meassure IO?.. in freebsd there is gstat[22:09:42] <Licenser> jeffpc I had a chance to look at your logfiles and for me it looks like something went wrong with the communication between your systems, sniffle and wiggle report timeouts for calls while chunter says the connection is refused[22:11:10] <jeffpc> Licenser: it's just one vmware guest... they should be able to talk to each other[22:11:29] <Licenser> I agree, perhaps when you installed chunter alone there was a typo ro something?[22:11:40] <killfill> its strange.. when i run the tests on centos myqsl process uses like 30% cpu.. on centos it uses 10% (as i can see on prstat)[22:12:32] <jeffpc> Licenser: don't think so... but it is possible[22:13:00] <jeffpc> Licenser: where are the packages stored? because the one package I had showed up[22:13:16] <jeffpc> Licenser: just no machines & images[22:13:29] <Licenser> the packages live in the redis database, that gets queried by snarl. Images and machines come from chunter[22:13:36] <nahamu> are the two database servers using the same backend?[22:13:36] <Licenser> do you have the system running/[22:13:51] <jeffpc> nope, it's on the laptop at home[22:13:53] * jeffpc is at work[22:13:57] <Licenser> okay[22:14:11] <nahamu> killfill: what backend is each mysql server using?[22:14:20] <killfill> nahamu: innodb[22:14:30] <Licenser> well when you get home try doing: /opt/chunter/bin/chunter stop and then/opt/chunter/bin/chunter console (that way it boots in the forground)[22:14:39] <jeffpc> ok[22:14:42] <nahamu> do they have identical configs modulo IP address?[22:15:21] <killfill> nahamu, modulo IP address?[22:15:56] <Licenser> The ouit actually looks all good o.O[22:15:56] <nahamu> sorry, is everything the same (or if not, is everything the same except for an IP address)?[22:15:56] <killfill> nahamu: well one has " default-storage-engine = innodb innodb_log_file_size = 400M" and the other not[22:16:07] <nahamu> which one has that?[22:16:10] <killfill> aah not reallt 100%.[22:16:21] <killfill> nahamu: ill setup the things again wich exactly the same version, config, etc.[22:16:25] <killfill> nahamu: smartos one[22:16:42] *** neophenix_ has joined #smartos[22:16:55] <Licenser> hmm hmm jeffpc your VM is a oddity :P[22:17:12] <jeffpc> glad to be of service :)[22:17:33] <Licenser> Sadly I\m not at home and have bearly intenet or I'd try it out myself :( if we don't get it fix I promise I'll do that once I'm back in berlin[22:17:39] *** ryancnelson has joined #smartos[22:17:52] <jeffpc> no worries[22:17:57] <Licenser> what might be best: just shutdown chunter and stop the zone, start it again then restart chunter[22:18:01] <Licenser> that should give you a pretty clean state[22:18:57] *** neophenix has quit IRC[22:18:57] *** neophenix_ is now known as neophenix[22:22:58] <killfill> oh hell.. sysadmin put a router between my smartos.. thats not fair :)[22:24:13] <killfill> its 2x faster.. :)[22:24:51] * killfill will stop floodding the channel now.. :P[22:33:09] <rmustacc> So it ended up being the network traffic?[22:36:37] *** denizr has quit IRC[22:36:44] *** denizr has joined #smartos[22:40:16] <killfill> rmustacc: yup, looks so. but ill build a less esoterilcal bench.. :P[22:50:40] *** denizr has quit IRC[22:59:29] *** denizr has joined #smartos[23:06:49] *** jamesd has quit IRC[23:07:43] *** jamesd has joined #smartos[23:14:58] *** CarlosC has quit IRC[23:50:15] <killfill> hm... it seems the cpu capping is not working right, or better that i understand.. :P[23:51:09] <killfill> i have a 2 smartos mashines. one with 1 vcpu and other with 4vcpu.. cpu performance is almost equal[23:51:54] <ryancnelson> vcpu settings aren't used at all in smartos machines[23:52:07] <ryancnelson> cpu shares and caps are what constrain zone resources[23:52:40] <ryancnelson> the vcpu settings are just in the package definitions in case you use that package to deploy VM's[23:53:42] <killfill> aah ill play with the cpu cap then.. :P[23:54:51] <ryancnelson> cap == "you shall go no higher than this" … shares == "this is your guaranteed minimum cpu, even if all other zones are busy"[23:55:21] <ryancnelson> … high caps mean that zones can burst up to most of the box's cpu, if everyone else is idle[23:55:31] <ryancnelson> which is kind of awesome, for sharing workload[23:56:07] <ryancnelson> we mostly cap so that people don't buy a 1gb machine and happen to land on a 90% empty host, and then get really used to having 24 cores to redline all day[23:56:26] <ryancnelson> … then complain when performance falls down to "merely awesome"[23:57:01] <killfill> hehe..[23:58:05] <killfill> so "30% burst" that is on the joyent public cloud is setgin up like: cpu share: equivalent to 1 cpu of the box, and cap to '130' ?