December 25, 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 | >
December 25, 2012 [00:00:02] <wesolows> it actually is now. that's what the change from 14.4 kbps to tens of Mbps has enabled.[00:00:20] <wesolows> when you have a good pipe, storing stuff somewhere else ceases to be a problem.[00:00:36] <lundh> yes but colo storage is expensive[00:00:43] <wesolows> ryancnelson: yeah, most home systems are hopelessly unbalanced[00:01:14] <wesolows> they have huge, excruciatingly slow and unreliable storage, puny CPUs with tiny caches, and not nearly enough DRAM for the storage they have.[00:01:25] <ryancnelson> yep. that pipe means that my "cold storage" can be at home, though, instead of the menagerie of various small servers, which is 100% flipped from what i used to have[00:01:50] <lundh> my home server is the other way around. way to fast for my needs[00:02:01] <ryancnelson> i used to have 6 servers at the house. now i have a mac and vmware, and an assload of vms "out there"[00:02:23] <ryancnelson> also a dresser drawer full of ide disks from days of yore.[00:02:37] <wesolows> I have an Ivy Bridge workstation... it's amazing how much faster a middle of the line Sandy Bridge server clocked 25% lower is.[00:02:54] <ryancnelson> one is a 10gb HD labeled "hugefriggindisk" … ah, 1997[00:02:57] <lundh> heh. I have a HTCP, a laptop and the server[00:02:58] <wesolows> makes any PC-class system look like a turtle.[00:03:22] <Zigara> wesolows: ivy brdige has 5-15% cpu performance increase[00:03:31] <Zigara> comapred to sandy[00:03:47] <wesolows> Zigara: yeah but the delta between PC-class and Real CPUs is a lot bigger than 15%[00:03:47] <Zigara> even higher with the gpu, I think2 5+[00:03:53] <wesolows> it's probably closer to 15x.[00:03:54] <lundh> Zigara: same architecture, should it not be exaclty the same at the same freq?[00:04:10] <wesolows> GPUs are completely useless.[00:04:11] <lundh> excep gpu that is[00:04:14] <Zigara> I'm not sure the details honestly[00:05:00] <lundh> wesolows: intel pretty much makes no difference between some of the xeons and i7[00:05:00] <wesolows> if you want to test this, take an ivy bridge PC system with an SSD and build illumos, then do the same on a sandy bridge server clocked 25% lower[00:05:17] <Zigara> wesolows: I didn't see you said server at first[00:05:23] <wesolows> there is absolutely no comparison[00:05:36] <lundh> wesolows: how do you define server? xeon cpu?[00:05:41] <wesolows> the server with plain spinning disks will blow away the PC[00:05:46] <Zigara> I used intel in all my servers and desktops, I don't have any ivy bridge servers yet though[00:05:54] <wesolows> I don't know the marketing names... E5-2670 for example.[00:06:08] <Zigara> I do have a sandy bridge box with a dual e5-2620 with 128gb of ram lol[00:06:25] <Zigara> I haven't seen any ivy bridge e5's though[00:06:31] <wesolows> they're not out yet[00:06:37] <Zigara> thought so[00:06:38] <lundh> wesolows: my guess is that the sandy bridge server has 25% more L3 cache[00:06:53] <wesolows> it's bigger, yes[00:06:54] <lundh> wesolows: it makes all the difference[00:07:00] <wesolows> but that doesn't account for all the performance difference[00:07:10] <Zigara> running this in my laptop: machdep.cpu.brand_string: Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz[00:07:13] <Zigara> ivy bridge based, been loving it[00:07:14] <lundh> it does, there is no other difference[00:08:23] <wesolows> yeah, sounds fast but it isn't.[00:08:37] <Zigara> sure it is[00:08:39] <wesolows> the memory bandwidth in the real systems is a big contributor[00:08:40] <lundh> clock freq, cache and "turbo boost" and extensions (if the sw can use them) is the only differences between *bridge cpus[00:09:40] <lundh> the only peformance difference that is[00:09:58] <lundh> ivy require less power[00:10:21] <wesolows> I guess. TBH I haven't explored all the differences quantitatively, I just know one builds about 3x as fast as the other.[00:10:44] <lundh> qual core compared to quad core? :)[00:11:11] <wesolows> the parallelism in our build isn't very good[00:11:43] <lundh> without seeing the hw specs its hard to say anything else :)[00:12:24] <wesolows> i5-3450S (QC, 2.8) vs E5-2670[00:12:53] <lundh> ran a compute cluster for a while with a few different cpus. the generation before *bridge-cpus. when all the environment were lined up the peformance scaled linearly with cpu freq[00:13:22] <Zigara> lol that e5 has 16 threads[00:13:30] <Zigara> and 20mb of cache[00:13:33] <wesolows> yes it does. but again, our build's parallelism sucks.[00:14:13] <Zigara> I'd have told you the e5 was better from the start[00:14:24] <wesolows> the 2670 is a nice CPU. it performs as I would expect given its specs. but the desktop CPUs are very disappointing relative to their specs.[00:14:28] <lundh> wesolows: I'd say the extra cores make the difference. lets the build system reside on a few cores uninterrupted[00:14:55] <lundh> not breaking the pipeline saves a lot of time[00:14:56] <wesolows> *shrug* maybe. it was not an investigation and I gathered no useful data. merely an anecdotal observation.[00:15:00] <Zigara> Recommended Customer Price BOX : $1556.00[00:15:07] <Zigara> vs $195[00:15:10] <Zigara> for the i5[00:15:21] <wesolows> yeah. and I think 8x is probably about what it's worth too.[00:15:40] <Zigara> I would never comapre those 2 cpus[00:16:12] <wesolows> I'm not comparing them in a "should I buy X or Y" sense at all. Just observing that the 8x price is well worth it in terms of performance.[00:16:36] <lundh> well, 8x price, 3x performance. seems like a bad deal[00:16:50] <Zigara> xeons are insane these days, but that being said, you get quite a huge amount of performance out of the latest desktop grade chips[00:17:03] <wesolows> it is if you have unlimited time to wait[00:17:40] <wesolows> I guess. it doesn't feel to me like the desktop chips have advanced at all in the last 10 years[00:17:45] <lundh> Zigara: thats the tricky part. you cant say xeons. it differes alot. some of the xeons are just rebranded i7 with ECC support instead of fast graphics[00:17:53] <wesolows> obviously that's subjective; I'm sure objectively they're much better.[00:18:06] <Zigara> lundh: well yes, it would have made my point much longer though[00:18:12] <lundh> :)[00:18:27] <Zigara> I'm very content with current gen desktop hardware[00:18:46] <wesolows> the E5s feel to me the way that the last alpha system I ever saw felt. just blew me away.[00:18:47] <Zigara> my desktop is running at 4.4ghz on stock air cooling, sandybridge, slightly outdated now[00:19:04] <Zigara> wesolows: E5's are beasts[00:19:14] <lundh> I dont have a desktop[00:19:15] <Zigara> I can't wait to see the ivy brdige version[00:19:30] <Zigara> bridge*[00:19:41] <wesolows> I don't really care about ivy bridge. Our ops guys might, though, if it really does use a lot less power.[00:19:56] <lundh> it doesnt[00:20:04] <wesolows> that's my impression as well.[00:20:09] *** wolfeidau has joined #smartos[00:20:18] <lundh> the difference is smaller then the math says it should have been[00:20:35] <wesolows> well, in part that's because the CPUs aren't the only thing using power[00:21:02] <lundh> the next tick is only a few months away, lets see what that brings[00:21:04] <wesolows> DRAM adds up, and disks consume a lot. all said, a 20% drop in CPU power would translate to probably 4 or 5% in total.[00:21:25] <lundh> sure but even the cpu figures are a bit off[00:21:39] <ryancnelson> if we cared about power savings that much, fixing the c-states thing would be a higher priority[00:21:45] <wesolows> sure; if it's actually 10% and not 20% you'd probably have a hard time measuring it at all[00:21:54] <lundh> exactly[00:21:59] <wesolows> ryancnelson: you mean the firmware bugs that cause systems in C6 to lock up or lose interrupts?[00:22:12] <wesolows> yeah, fuck that shit. C6 is an abomination.[00:22:27] <lundh> wesolows: that being said, the LP-versions are pretty nice if you can take the performance penalty[00:22:37] <wesolows> the solution to that is to increase tenancy, not dick around with shoddy broken firmware[00:23:37] <wesolows> if any CPU would ever be in C6, we're already throwing away a lot more money than just the marginal power it dissipates in C3[00:24:24] <lundh> wesolows: just to check. nfs share from the GZ to VMs and Zones should be ok? resharing a NFS share through AFP should also be fine?[00:24:42] <wesolows> I... think so. I haven't tested it myself.[00:25:29] <lundh> that could be my ticket back into my confort zone while I look into the process you proposed[00:25:36] <wesolows> we never supported AFP at fishworks either so I'm kind of in the dark there.[00:25:55] <lundh> I can test that[00:26:09] <wesolows> if you never try to use file/record locking to manage simultaneous writes, it'll probably be fine.[00:26:32] <lundh> dont think that I do[00:26:57] <wesolows> like, I have zero idea what happens if you use AFP to try to lock [0x1000,0x2000) for writing from one client while you try to lock that same range from an NFS client.[00:27:22] <wesolows> or if AFP even supports locking at all.[00:27:25] <lundh> thats exactly what I'm worried about[00:28:34] <wesolows> well, one thing's for sure: you'd have this problem no matter what OS you use, if you use this architecture.[00:28:41] <wesolows> so you may as well try it and see.[00:28:47] <lundh> I know[00:30:25] <lundh> but its new territor for me so just being able to discuss it helps alot[00:40:39] *** ryancnelson has quit IRC[00:44:48] *** masked has quit IRC[00:49:49] *** newlix_ has joined #smartos[00:51:04] *** masked has joined #smartos[00:51:04] *** masked has joined #smartos[00:54:25] *** newlix_ has quit IRC[01:04:17] <lundh> weird, trying to build transmission. the configuration fails at openssl of all things[01:12:02] <AlainODea> Is there a performance disadvantage to using a 32-bit image like base vs a 64-bit image like base64?[01:18:55] <masked> i guess it would require the use of additional compat libs[01:19:39] *** Vod has joined #smartos[01:21:57] <konobi> AlainODea: memory usage[01:27:34] <AlainODea> konobi: as in lower memory usage on 32-bit zones/VMs, is that correct? I know Java tends to be better off in 32-bit because of its garbage^2 collector[01:29:38] *** daleg has quit IRC[01:32:28] <konobi> AlainODea: yeah, 32 bit binaries/libs[01:39:32] <AlainODea> konobi: cool, no deeper than I thought. I was worried there was a specific performance penalty for emulating 32-bit.[01:43:14] <wesolows> there's no emulation required[01:43:55] <AlainODea> wesolows: good point. It's a zone. I have to unwire that assumption.[01:43:57] <wesolows> the 64-bit kernel has always been able to run 32-bit user processes directly on metal; support for the 32-bit ABI is built into the hardware and is fully supported by the kernel[01:44:03] <wesolows> yep[02:09:05] <AlainODea> Should I be concerned about this warning? "WARNING: Couldn't read ACPI SRAT table from BIOS. lgrp support will be limited to one group"[02:09:40] <AlainODea> I get it on a Dell PowerEdge R720xd with 1.3.6 BIOS (latest available)[02:11:55] <konobi> nope[02:13:26] <wesolows> if you have only 1 CPU, it's a non-issue[02:13:41] <wesolows> if you have more than one, your system firmware is buggy and performance may be reduced as a result[02:13:50] <wesolows> it does not affect correctness in either case[02:14:34] <wesolows> (actually, your system firmware is buggy regardless, but with only 1 CPU there's no difference at all)[02:20:02] *** masked has quit IRC[02:20:57] *** Vod has quit IRC[02:28:00] *** jamesd has quit IRC[02:40:40] <AlainODea> wesolows: it has dual 6-core Xeon E5-2640 @ 2.50GHz in it. Nothing exotic, but it's quite powerful.[02:41:14] <AlainODea> wesolows: any troubleshooting steps I can take to see where this fails?[02:41:23] <wesolows> so it does have 2 populated sockets? if so, the missing or busted SRAT table is bad. your system will work, but not as fast as it otherwise might. File a bug with Dell.[02:41:28] <wesolows> nope.[02:41:35] <wesolows> your firmware is busted.[02:41:52] <wesolows> and the defining attribute of firmware is that it's completely unobservable.[02:46:21] <AlainODea> wesolows: I got the same thing with a Dell Precision T3500.[02:47:17] <konobi> might be a bios option[02:48:57] <wesolows> yeah, that's possible.[02:49:21] <wesolows> you may be able to use tools like iasl (?) to dump the acpi tables[02:50:13] *** newlix_ has joined #smartos[02:50:57] <AlainODea> konobi: fun times. The BIOS on these machines has some annoying limitations and is intended to be navigated with a mouse. Apparently the irony of this is lost on Dell ;)[02:51:32] *** yofuh has quit IRC[02:51:37] <konobi> yeah... wesolows may remember my eternal anger at the bios on the dell r210s[02:51:40] *** Vod has joined #smartos[02:52:02] *** yofuh has joined #smartos[02:53:58] <wesolows> all BIOSes are terrible, and Dell's are especially egregious[02:54:11] <wesolows> the BIOS should have been left in 1987 where it belonged.[02:54:34] <konobi> i should have a play with coreboot again[02:54:40] <wesolows> but Dell and others want you to run Windows, because Microsoft pays them to, so they keep this ancient stuff that Windows relies on.[02:54:42] *** newlix_ has quit IRC[02:55:02] <konobi> the r11 and r12 stuff is all EFI now[02:55:03] <wesolows> coreboot is fine, works a champ in QEMU. But support for decent HW is 2-5 years behind.[02:55:26] <wesolows> EFI is even worse[02:55:32] <konobi> qemu just uses seabios directly now[02:55:41] <wesolows> you can build coreboot without seabios :-)[02:55:49] <konobi> yup[02:56:11] <wesolows> and you can point qemu at your own coreboot image too.[02:56:16] <konobi> so it's EFI to configure it, but it's in bios legacy mode... >_<[02:56:38] <konobi> wesolows: yeah... was playing with getting OSX booting on smartos a while back... fun fun[02:56:48] <wesolows> the solution to the problems with the BIOS was not to replace it by something even bigger and more complex like EFI, it was to get rid of it completely.[02:57:01] <wesolows> sigh.[02:57:42] <konobi> well, uboot is a pain too[02:58:00] <wesolows> there's a better way than that too[02:58:17] <konobi> mmm?[02:58:20] <wesolows> I'm out to prove it. More to come![02:59:00] <konobi> ruh-roh[03:00:06] <AlainODea> wesolows: hating this BIOS labyrithe. I expect I'll bump into David Bowie shortly...[03:00:26] <wesolows> yeah, sorry. it sucks. complain to dell, or just don't buy their stuff.[03:00:59] *** Vod has quit IRC[03:01:08] <konobi> it all sucks![03:01:43] <konobi> got an indication of how much it'd cost to get a SAS HBA made up recently[03:02:43] <wesolows> oh yeah?[03:03:09] <AlainODea> wesolows: Dell's a Joyent partner... Peronsally I'm incline to go SuperMicro. I think I found the setting. There is CPU Power Management under System Profile Settings. It lets me choose between System DBPM (DAPC), Maximum Performance and OS DPBM. Am I right in assuming I want OS DPBM when using SmartOS?[03:03:35] <wesolows> I have no idea what those are.[03:03:47] <wesolows> nor how they'd affect the SRAT table.[03:04:24] <konobi> wesolows: yeah, wasn't too bad actually... but i got the details on the down low from someone with an NDA, so can't really say[03:04:48] <konobi> acpi?[03:05:04] <wesolows> yeah, the problem is the ACPI SRAT table[03:05:19] <wesolows> but I don't know what DAPC, DBPM, etc. are in this context.[03:06:31] <wesolows> anyway, I've gotta run. Hopefully the manual will help! :-/[03:06:51] <konobi> node interleving?[03:08:28] <AlainODea> wesolows: take care and thank you for your help :)[03:08:37] <konobi> you probably want to disable node interleaving[03:09:03] <AlainODea> konobi: I'll take a look. Any keywords I should look for in the BIOS?[03:09:30] <konobi> that's how i read it from the r720 manual[03:10:53] <konobi> AlainODea: https://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&ved=0CEoQFjAB&url=http%3A%2F%2Fsupport.dell.com%2Fsupport%2Fedocs%2Fsystems%2FpeR720%2Fen%2FOM%2Fr720omen.pdf&ei=5wnZUKX1NK7oiAKpyoDIAQ&usg=AFQjCNFjjDOsJK5jmch-1npZ2STV_S5Gxw[03:11:55] <AlainODea> konobi: thank you. I really should reach for the manuals in this situation. That or the support we pay handsomely for from Dell.[03:12:13] <AlainODea> konobi: I'll do some reading of that and see where I get. Thanks again.[03:15:15] <konobi> np[03:19:01] *** jamesd has joined #smartos[03:19:12] *** miine__ has joined #smartos[03:22:10] *** miine has quit IRC[03:22:11] *** miine__ is now known as miine[03:24:34] <AlainODea> konobi: I don't see the link between ACPI SRAT and Node Interleaving in the manual, but it is mentioned elsewhere. Where did you learn this? I'm trying to learn to fish here so I stop coming and asking for supper here on IRC :)[03:26:17] *** Vod has joined #smartos[03:26:46] <konobi> AlainODea: i've seen it elsewhere[03:26:59] <konobi> srat has links to numa, etc.[03:28:50] <AlainODea> konobi: I have so much to learn on this. I started in Java programming, added Linux Administration, added IP Networking, added Windows Administration. That doesn't touch the other programming languages and frameworks I have rattling around. My head is close to exploding :)[03:30:35] <AlainODea> konobi: it is difficult to be more than a novice at such a broad knowledge set. I wish there were less knowledge required, but yet that's why this is actually a way to make money.[03:31:15] <AlainODea> konobi: not that money is why I do this, because it certainly isn't.[03:33:50] *** Vod has quit IRC[03:41:06] <konobi> heh, i need to stop giving away free advice ^_^[03:49:52] <AlainODea> konobi: LOL, indeed. However, you may have won yourself a great reference. I'm researching SmartOS now, but if Verafin buys in I will strongly recommend that we look to Cloudtone for depth in support :)[03:50:43] *** masked has joined #smartos[03:50:44] *** masked has joined #smartos[03:51:11] <AlainODea> konobi: your solution of disabling Node Interleaving worked. The ACPI SRAT warning is gone. Very likely the same reason it isn't working on my T3500 on my desk :)[03:53:32] *** Webhostbudd has joined #smartos[03:56:00] *** masked has quit IRC[04:06:08] <konobi> oh, nice[04:14:47] <konobi> NL?[04:24:30] *** dysinger has quit IRC[04:34:40] <AlainODea> konobi: Yes, we're in St. John's, NL[04:35:06] <AlainODea> konobi: http://verafin.com/contact[04:36:33] <konobi> ah, more canucks, heh[04:38:15] <AlainODea> konobi: indeed. In fairness to my colleagues we have an office in Alabama with some top-notch folks as well. The majority of R&D and Support is here is SJ though.[04:38:26] <AlainODea> konobi: *in[04:48:09] <konobi> aren't you supposed to be able to fish from the womb?[04:49:30] <AlainODea> konobi: touché[04:51:11] <konobi> *rimshot*[04:51:26] <AlainODea> konobi: "Give a person a fish and they'll eat for a day. Teach that person to fish and they'll eat for a lifetime." I don't know whose quote it is, but one of favourite professors, Dr. Dennis Peters, said that to our class one day and it shaped the way I live and learn.[04:52:13] <AlainODea> konobi: I've gotta jet. It's late here and there's a busy day tomorrow with four kids in tow :) Take care.[04:53:25] <konobi> ttfn![05:26:00] *** masked has joined #smartos[05:26:00] *** masked has joined #smartos[05:26:10] *** dysinger has joined #smartos[05:42:09] *** Derzzle has joined #smartos[05:43:14] *** Derzzle has quit IRC[06:58:58] *** dysinger has quit IRC[07:37:20] *** masked has quit IRC[08:09:57] *** cjones_ has quit IRC[08:22:05] *** cjones_ has joined #smartos[08:22:24] *** cjones_ has joined #smartos[08:44:20] *** jamesd has quit IRC[08:56:30] *** Cpt-Oblivious has joined #smartos[09:05:35] *** Webhostbudd has quit IRC[09:23:26] *** texarcana has quit IRC[09:24:38] *** texarcana has joined #smartos[09:27:56] *** sebasp__ has quit IRC[11:24:45] *** sebasp__ has joined #smartos[12:37:37] *** darjeeling has quit IRC[12:37:45] *** darjeeling has joined #smartos[12:47:45] <lundh> is is not possible to have users with names longer then 8 characters?[12:58:15] *** xmerlin has quit IRC[13:24:47] *** Livid has joined #smartos[13:25:49] <Livid> How can I try SDC on my own machine? The adoption page has nothing to download[13:26:31] <MerlinDMC> Livid, is the adoption page back online?[13:27:04] <Livid> I can't find anything meaningful on adoption.joyent.com[13:27:27] <konobi> the adoption program is no longer available[13:27:32] <MerlinDMC> yeah ... that's the normal jyoent homepage ... the adoption program formular is gone for some months now[13:28:04] <Livid> So how can I get SDC?[13:28:28] <MerlinDMC> i would try to contact joyent sales and ask for it ...[13:29:33] <Livid> Oh thanks. But I'm not ready to buy it, now I just want to give it a try on several internal servers.[13:30:25] <MerlinDMC> maybe sales can give a limited trial away also ... i don't know - but asking mostly never costs anything[13:30:35] <konobi> you'd have to talk to pre-sales[13:30:48] <Livid> Sure, thanks![13:32:04] <Livid> I watched the install video on YouTube, SDC is quite impressive. Why Joyent didn't open it just like OpenStack?[13:32:46] <konobi> as far as I can tell openstack is a gong show[13:33:03] <lundh> is there a simpel way to compile in oen zone and them move the binaries to another one?[13:33:58] <lundh> I want to get netatalk 3.0.1 installed but I dont want the build-tools on my file server[13:34:29] <MerlinDMC> lundh, build a targz package and copy it ... or copy the folder under /opt ...[13:34:36] <Livid> Yes it's a pain to deploy OpenStack, and SDC is quite ready for business.[13:35:30] <lundh> MerlinDMC: building 101: how do I build a targz?[13:35:59] <MerlinDMC> "tar c" ? :)[13:36:34] <MerlinDMC> if you're using pkgsrc just run a "make package" i think[13:36:51] <lundh> I'm not[13:37:12] <lundh> building from a tarball downloaded from the netatalk website[13:37:42] <MerlinDMC> then just build with a nice --prefix=/opt/netatalk and create a tar archive of the directory[13:38:27] <lundh> how do I handle SMF manifests in that case?[13:51:31] *** Livid has quit IRC[13:53:48] *** jelmd has joined #smartos[14:43:55] *** yofuh has left #smartos[14:44:13] *** yofuh has joined #smartos[15:14:03] *** jamesd has joined #smartos[16:10:57] *** Cpt-Oblivious has quit IRC[16:34:00] *** xmerlin has joined #smartos[16:52:11] *** newlix_ has joined #smartos[16:56:52] *** newlix_ has quit IRC[17:02:48] *** jelmd_ has joined #smartos[17:02:48] *** jelmd has quit IRC[17:02:56] *** jelmd_ is now known as jelmd[17:06:35] *** linuxprofessor has quit IRC[17:11:48] *** linuxprofessor has joined #smartos[17:19:54] <linuxprofessor> hmm… still getting "timed out waiting for /var/svc/provisioning to move…", no ideas yet on how to solve it? =)[17:24:18] <konobi> svcs?[17:29:32] <linuxprofessor> i'm guessing that wasn't in response to what i wrote? hehe[17:29:32] *** Seze has quit IRC[17:30:08] <konobi> no, seriously[17:30:28] <linuxprofessor> [root@00-15-17-19-34-c8 /zones/json_files]# vmadm create -f europa.json[17:30:29] <linuxprofessor> timed out waiting for /var/svc/provisioning to move for ff010e2b-d986-47b7-9ae2-20b9ef6cbcc9[17:30:47] <xmerlin> linuxprofessor, same here ...it was a bad parameter in the payload[17:31:19] <xmerlin> bad ip in my case[17:31:27] <konobi> linuxprofessor: `svcs`, just want to make sure everything is good[17:31:35] <linuxprofessor> oh, ok =)[17:31:45] <linuxprofessor> xmerlin: i'll have another look[17:32:01] <konobi> xmerlin: you also had your netmask bad, iirc[17:32:23] <xmerlin> konobi, no only bad ip[17:32:30] <xmerlin> bad cut and paste ..[17:32:35] <konobi> well, your gateway was outside of your netmask at least[17:32:38] <xmerlin> a 6 at the end[17:32:51] <xmerlin> yes[17:32:59] <xmerlin> because I've addeded a 6 at the end[17:33:03] *** jelmd has left #smartos[17:33:12] <konobi> ah[17:33:13] <xmerlin> s/addeded/added/[17:35:54] <xmerlin> could you point me a good ipfilter guide? ...the performances of pfsense 2.1x are not so good ...and I need to write firewall rules inside a smartos zone[17:39:51] <konobi> freebsd ipf docs are great[17:40:06] <xmerlin> ok[17:45:07] <xmerlin> as I can see tcpdump doesn't work as expected ...is it a known issue?[17:45:37] <konobi> you can use snoop[17:46:26] <xmerlin> ok[18:28:29] <xmerlin> I've found a very strange behaviour in a VM ...I cannot boot from a cdrom "boot failed: could not read from cdrom (code 0003)" ...same image works fine on another VM[19:26:21] *** gkyildirim has joined #smartos[19:32:20] *** stev_ has joined #smartos[19:38:33] <Alasdairrr> Are there any sngl datasets available for download?[19:42:26] <wesolows> no[19:43:38] <Alasdairrr> I wonder how aszeszo got his sngl zone up. Hmm.[19:44:22] <Alasdairrr> I guess I can take an existing dataset and fudge it into shape manually[19:45:07] <wesolows> um, maybe. given that it's nowhere near ready to ship, hasn't even been brought up widely on lab systems, and undoubtedly has dozens of serious issues left to work out, I wouldn't bet on it[19:45:30] <Alasdairrr> I'm happy experimenting[19:45:44] <Alasdairrr> I just need to get it vaguely up so I can start building software under /usr[19:45:56] <wesolows> sure... just be sure not to set expectations based on a preliminary idea of what it may or may not someday become[19:46:16] <Alasdairrr> You don't need to slap safety warnings on it, I'm a big grown up :-)[19:46:43] <Alasdairrr> If it deletes my system, renders me impotant, gives me cancer, I won't complain[19:46:54] <wesolows> I'm not worried about you having a bad experience; I'm worried about people doing stuff that is based on assumptions about it, when we haven't committed to interfaces.[19:47:15] <wesolows> and then being angry/upset/creating unnecessary forks based solely on expectations that never should have existed[19:48:13] <Alasdairrr> We maintain our own fork of SmartOS anyway, for internal use, with our own illumos patches and smartos patches. If SNGL doesn't meet our needs we'd probably create our own brand anyway, it was kind of my intention to look into it at some point.[19:48:33] <wesolows> ok[19:48:37] <konobi> creating the joyent brand was kinda fun[19:48:38] <wesolows> good luck then![19:48:40] <Alasdairrr> But literally all I need is a filesystem layout, I could do this in a chroot[19:48:47] <Alasdairrr> Thanks :-)[19:48:55] <Alasdairrr> In fact it might be quicker/easier to do this in a chroot, hmm[19:48:58] * Alasdairrr strokes his beard[19:56:01] <nahamu> what exactly is sngl (or what will it be when it's fully baked)?[19:56:48] <Alasdairrr> It's basically a writable /usr, with very little under /usr such that the end user can shove their own software in there[19:56:58] <Alasdairrr> as I understand it, anyway[19:57:18] <Alasdairrr> From what I gather the intention is to have pkgsrc in an sngl zone under /usr, with config under /etc and var stuff under /var[19:57:28] <Alasdairrr> So locations are entirely normal and what end users expect[19:57:55] <nahamu> and th zone is then pretty completely separate from the plaform.[19:58:20] <Alasdairrr> it'll still be "sparse" in that the system /usr gets mounted under /system/usr read-only[19:58:44] <Alasdairrr> So a smartos upgrade doesn't require updating software in the zone[19:58:49] <nahamu> ah, but the point is that th platform /usr will go under /system/usr[19:58:53] <Alasdairrr> yeah[19:59:43] <nahamu> interesting.[20:00:05] <Alasdairrr> it is, very[20:00:08] <nahamu> I don't think I need anything like that, so I'll probably wait until it's fully baked. :)[20:00:29] <Alasdairrr> We'll be populating /usr with ec-userland (http://smartos.pkg.ec/)[20:02:53] <nahamu> will the fact that you can put things in /usr simplify ec-userland?[20:03:32] <nahamu> or is it more that your end customers will find /usr less confusing than /ec?[20:03:37] <Alasdairrr> The latter[20:03:49] <Alasdairrr> I've never liked /ec really, it probably stops people in the wild using it[20:04:09] <Alasdairrr> We're going to promote it and offer support on it for people when sngl becomes a supported brand[20:04:26] <nahamu> cool[20:04:55] <Alasdairrr> It's basically a stack of software built specifically for web hosting, we're a managed hosting company[20:05:28] <nahamu> and in theory the sngl brand could run an EC IPS dataset, or a Joyent pkgsrc one, etc.[20:05:35] <Alasdairrr> precisely[20:05:49] <Alasdairrr> If Joyent don't make it depend on pkgsrc too much[20:11:44] <nahamu> I'd think it would be most flexible if it doesn't depend on pkgsrc at all...[20:12:08] <Alasdairrr> I don't see any reason why it would[20:13:44] <nahamu> oh, the brand stuff as already (started to?) land in illumos-joyent...[20:14:05] <Alasdairrr> Yup![20:14:27] <Alasdairrr> There are no instructions and no datasets, I've shelved actually using it for now, I'm just setting up a chroot that emulates what i'd get with sngl[20:14:45] <Alasdairrr> unfortunately /sbin/zfs won't work even though i've crle'd in /system/usr/lib[20:15:00] <Alasdairrr> although[20:15:06] <Alasdairrr> that might be due to missing doors and what not[20:15:09] <Alasdairrr> rather than due to libraries[20:15:45] <nahamu> does it start and then freak out, or downright refuse? could truss help?[20:16:43] *** jelmd has joined #smartos[20:16:50] *** jelmd has left #smartos[20:17:26] <Alasdairrr> just trying truss now[20:17:34] <Alasdairrr> got proc mounted up now[20:18:02] <Alasdairrr> I don't really expect this stuff to work in a chroot anyway really[20:19:42] <Alasdairrr> 3887: stat64("/usr/lib//libshare.so.1", 0x08046CAC) Err#2 ENOENT[20:20:38] <Alasdairrr> Guess those things don't use the dynamic runtime linker search path due to their use during boot[20:26:00] <nahamu> there's got to be some joke to be made about a Unix tool needed for boot that's hardcoded to look in /usr, but the audience for it is probably rather small.[20:26:27] <Alasdairrr> lol[20:26:56] *** gkyildirim has quit IRC[20:27:25] *** dysinger has joined #smartos[20:30:31] <wesolows> stat64 instead of dlopen? naughty.[20:31:00] <wesolows> that's probably a Fishworks remnant though... I believe we made libshare optional for ZFS, so that shouldn't be fatal.[20:31:48] *** xmerlin has quit IRC[20:35:49] <nahamu> so there should be other failures in the truss output?[20:44:29] *** xmerlin has joined #smartos[20:44:47] *** xmerlin has joined #smartos[20:48:08] <Alasdairrr> I will take a look[20:48:17] <Alasdairrr> just fighting with bash[20:51:05] *** dysinger has quit IRC[20:56:07] <jperkin> I _think_ that http://pkgsrc.smartos.org/packages/SmartOS/bootstrap/bootstrap-2012Q3-sngl.tar.gz is the latest used for sngl testing, of course it contains pkgsrc bits but it also has enough symlinks etc for boot to complete.[21:03:37] <xmerlin> I'm becoming crazy with the cdrom of a vm "could not read from cdrom (code 0003)" everytime ...it works fine in other vms[21:09:23] <wesolows> well, who is emitting the error?[21:09:50] <wesolows> once you know that, you know who to dtrace, which will give you a stack, and an errno, and ...[21:11:06] <xmerlin> bios via vnc[21:11:38] <xmerlin> could it be related to the fact I've 3 disks on this vm?[21:12:02] <xmerlin> I cannot see any other difference between this vm and the other vm where it works[21:12:40] <nahamu> xmerlin: is it an iso file? did it get properly copied to the root of the other VM's zone?[21:12:54] <xmerlin> yes it's an iso file of sysresccd[21:12:57] <xmerlin> yes[21:13:02] <xmerlin> I've copied it many times[21:13:26] <xmerlin> before addind the additional disks all works as expected[21:13:29] <xmerlin> adding[21:13:30] <konobi> but specifically into the zones/<uuid>/root folder[21:13:36] <xmerlin> yes[21:13:55] <konobi> is the dataset full?[21:14:34] <nahamu> if the dataset were full I think QEMU would fail to launch, but I could be mistaken on that one.[21:14:52] <xmerlin> zones 444G 162G 282G - 36% 1.00x ONLINE -[21:15:14] <konobi> dataset, not zpool[21:15:17] <xmerlin> do you mean the zone of the vm?[21:15:17] <nahamu> xmerlin: I think konobi meant zones/uuid[21:16:05] <xmerlin> zones/7134dac4-1eab-4d3c-accb-aa81440ae52c[21:16:05] <xmerlin> 10G 345M 9.7G 4% /zones/7134dac4-1eab-4d3c-accb-aa81440ae52c[21:21:57] <Licenser> morning![21:22:09] <konobi> xmerlin: virtio vs ide?[21:27:23] *** Licenser has quit IRC[21:28:46] *** dysinger has joined #smartos[21:29:42] <Alasdairrr> Random, "/usr/bin/which" is a csh script[21:31:04] <jperkin> indeed, yet bourne-and-derivative users still insist on using instead of type/command :)[21:32:05] <xmerlin> konobi, all virtio (disk/net) ...and cdrom ide but I've tried also virtio[21:32:50] <wesolows> don't forget hash![21:33:08] *** Licenser has joined #smartos[21:33:50] <Alasdairrr> hash? type/command?[21:34:00] <wesolows> help type, help hash, help command[21:34:21] <Alasdairrr> interesting, thanks wesolows[21:34:34] <Alasdairrr> I'm one of these "kids" that came to all this unix stuff via Linux. Slackware Linux specifically, back in 1996 or so[21:34:42] <wesolows> bash, when you're not touching its buggy bits, is underappreciated[21:34:57] <wesolows> but, oh, the bugs...[21:34:58] <Alasdairrr> and I have very lame habits built up over the years[21:35:09] <Alasdairrr> I only discovered "cd -" this year, most useful command I've learned in quite some time[21:35:15] <Alasdairrr> byebye pushd/popd[21:38:20] <Licenser> Alasdairrr it's one great command yap[21:38:41] <konobi> esc . is also a good one[21:45:00] <Licenser> for the record I lvoe ZFS[21:46:02] <nahamu> I think I would sleep much worse if my bits weren't on ZFS.[21:46:16] <Alasdairrr> I think we all love ZFS[21:46:35] <konobi> i've had linux shit itself twice in the last 7 days[21:47:29] <Alasdairrr> Almost all our clients wanted linux, and we thrust Solaris/SmartOS at them, and put up with it. I'm really looking forward to getting this sngl stuff up and running so I can make things even more familiar for them.[21:48:13] <konobi> all i would need is lxproc =0)[21:49:15] <nahamu> how do you turn on lxproc for a zone?[21:49:26] <nahamu> or is it just always on?[21:50:28] <Alasdairrr> It's always on with the joyent brand, under /system/lxproc[21:55:10] <wesolows> my long-term (personal) hope is that sngl will be a success, in that it will allow us to rid the joyent brand of abominations like lxproc[21:55:34] *** dysinger has quit IRC[21:55:35] <wesolows> consolidate all the horrid sewage in sngl[21:56:35] *** dysinger has joined #smartos[21:56:39] <Alasdairrr> haha[21:56:48] <Alasdairrr> Just create a sunos brand for yourself :-p[21:56:59] <wesolows> what's the difference?[21:57:13] <Alasdairrr> No lxproc, /usr/bin at the front of your path[21:57:17] <wesolows> if the GNU/Linux people all use sngl, those of us who prefer the OS as God intended can just use joyent.[21:57:26] <e^ipi> what's sngl ?[21:57:53] <nahamu> okay, if e^ipi didn't know what sngl is, I feel better for not having known what it was until today.[21:57:55] <wesolows> e^ipi: something that jperkin has posted a dataset for earlier today. experimental stuff I was trying to getpeople not to use.[21:58:20] <e^ipi> well what's it supposed to do?[21:58:33] <wesolows> someone said it kicks your dog in the head.[21:58:48] <wesolows> someone else suggested it will break your computer.[21:59:01] <nahamu> e^ipi: it lets you put whatever you want in /usr (which will then break the computer...)[21:59:15] <nahamu> minus the stuff in parentheses[21:59:24] <jperkin> e^ipi: pkgsrc built with prefix=/usr, system stuff mounted under /system with symlinks for essential bits, essentially.[21:59:27] <wesolows> probably causing you to accidentally kick your dog in the head as you miss the computer[21:59:56] <e^ipi> oh[22:00:14] <e^ipi> well yeah, that's going to break in fun & exciting ways[22:00:31] <wesolows> yep. the challenge is separating the deliberate breakage from the unintended[22:00:33] <e^ipi> seeing as how almost everything assumes uname == sunos != gnu[22:02:25] <jperkin> yep, I'm sure pkgsrc only did so well because it hides a lot of the breakage.[22:02:35] <wesolows> more specifically, most stuff out there assumes uname == sunos => it is 1993 and I am on Solaris 2.3 on a SPARC[22:03:20] <e^ipi> true[22:03:27] <Alasdairrr> you say that, but I find most stuff assumes Linux and breaks when /bin/sh isn't bash or whatever[22:03:37] <konobi> or dash[22:03:39] <konobi> *shudder*[22:03:48] <Alasdairrr> and by "most" I mean the fast-moving multimedia stuff like ffmpeg, vlc, etc[22:03:52] <wesolows> that's ok though; I've submitted a bunch of patches to make uname == linux => it is 1993 and I am using libc4 on Yggdrasil Plug-n-Play Linux 0.91.[22:04:02] <wesolows> hopefully that will help people understand[22:04:53] <wesolows> when they get a bunch of static COFF binaries they can't run, they'll learn what they've subjected the world to all these years[22:05:21] <Alasdairrr> jperkin: do you know what stuff under /usr/include gets symlink'd? In my fake snoogle chroot I've just symlink'd everything under /system/usr/include into it[22:05:28] <jperkin> Alasdairrr: everything[22:05:41] <Alasdairrr> cool[22:08:15] <nahamu> jperkin: what does sngl stand for?[22:08:38] <Alasdairrr> SmartOS's not gnu/linux[22:08:48] *** stev_ has quit IRC[22:08:54] <jperkin> what Alasdairrr said, pronounced "snuggle"[22:09:19] <nahamu> okay, that's pretty entertaining.[22:09:29] <Alasdairrr> It is :-)[22:09:30] <Alasdairrr> Very circular[22:09:40] <jperkin> but seriously, I can't stress enough, it's a really fugly hack only intented for very specific use-cases, so unless you are Alasdairrr and are ripping it to pieces for your own purposes you should probably just forget about it :)[22:09:57] <Alasdairrr> It's honestly not that hacky at all[22:09:57] <wesolows> +1 to that[22:10:03] <nahamu> I'm not touching it. I just like to understand the commits that scroll by.[22:10:03] <Alasdairrr> You guys are so boring[22:10:15] <wesolows> some of us care about things working[22:10:17] <wesolows> some don't[22:10:19] <wesolows> to each his own[22:10:26] <Alasdairrr> :-)[22:11:37] <nahamu> still, I would find it entertaining to live in a world where much of "the cloud" is "cuddle-licensed" "snuggle-zones"[22:11:55] <jperkin> the main benefits it has for me is to bring the prefix=/usr support in pkgsrc up to date, it's languished a little since a bunch of folks made it work for pkgsrc-based linux distributions they wanted.[22:12:59] *** Webhostbudd has joined #smartos[22:23:07] *** Cpt-Oblivious has joined #smartos[22:23:39] <lundh> a lofs interface in a zone, does that add a performance penalty compared to a delegated area?[22:25:55] <wesolows> yes, but it's unlikely to be measurable[22:26:05] <lundh> ok[22:26:14] <wesolows> *everything* comes with a performance penalty[22:26:29] <lundh> sure :)[22:57:01] <Licenser> hmm I've a very odd problem I try to create a zone and vmadm keeps timing out when I create new zones and I don't find any helpful logs indicating what's wrong :( any hint where I can start tracking the problem down?[22:57:18] <Alasdairrr> that definitely needs a FAQ section[22:57:23] <Alasdairrr> (if there isn't one already)[23:01:05] *** dysinger has quit IRC[23:02:10] *** dysinger has joined #smartos[23:45:40] *** Cpt-Oblivious has quit IRC