[05:42] *** Nahual has joined #arpnetworks [09:40] *** ameise has joined #arpnetworks [09:40] *** mkb_ has joined #arpnetworks [09:40] *** milki_ has joined #arpnetworks [09:42] *** mhoran_ has joined #arpnetworks [09:42] *** ChanServ sets mode: +o mhoran_ [09:42] *** karstensrage_ has joined #arpnetworks [09:43] *** mhoran has quit IRC (*.net *.split) [09:43] *** ant has quit IRC (*.net *.split) [09:43] *** karstensrage has quit IRC (*.net *.split) [09:43] *** mkb has quit IRC (*.net *.split) [09:43] *** milki has quit IRC (*.net *.split) [09:43] *** ameise is now known as ant [09:53] *** mkb_ is now known as mkb [10:21] *** karstensrage_ is now known as karstensrage [10:22] *** karstensrage has quit IRC (Changing host) [10:22] *** karstensrage has joined #arpnetworks [15:01] ix33: can you ping6 anything? [15:46] nathani: garry got it [15:47] it was a setting on a new thunder instance [15:47] VLAN [15:47] now, if someone would just fix vioblk on openbsd... [16:12] ix33: openbsd's disk subsystem in general isn't great :( [16:13] it really needs an overhaul but i haven't seen anyone looking to spend a lot of time on it [16:13] no kidding [16:13] untar of src.tar.gz on a physical 2-core box w/ssd in openbsd is ~100s [16:14] it's kind of embarassing, like when openbsd had no smp :) [16:14] which is awful [16:14] with or without softdep? [16:14] good question [16:14] i think slow untar with openbsd is to do with poor write behind [16:15] nosoftdep is default, so that [16:15] i'd try comparing with sofdep then [16:15] well that's a physical host [16:16] same default setup on a VPS instance is ~600s [16:16] ssd's aren't instant for acknowledging writes normally [16:16] they're faster than hard-disk, but still.. [16:16] if there are a lot of files it can add up [16:17] so if openbsd can keep writing stuff without waiting then performance can e higher [16:17] openbsd 6.0 src.tar.gz ? [16:18] mercutio: yes [16:18] i want to compare [16:18] i know that's a ton of little, highly compressed files... [16:18] yeah [16:19] going to put freebsd on this thing just to compare & feel depressed about it [16:20] i mean, 100s on the host i'm referring to is absolutely abysmal [16:24] i got 1 minute 2 seconds on a vps [16:24] with softdep though [16:25] yeah rebooting to try that [16:25] (/) [16:25] theo doesn't make fun of people for using softdep, does he? [16:25] no idea [16:26] i remounted with softdep [16:26] it seems you can do it live, at least it doesn't complain [16:27] -o remount works? [16:27] mount -u -o softdep / [16:27] 1 minute 50 seconds withotu softdep on vps [16:27] kvm/vioblk? [16:28] yeah [16:28] which version of openbsd? [16:28] 6.0 [16:28] something doesn't add up [16:28] how so? [16:28] you weren't using softdep [16:28] i got 1 minute 50 without softdep, 1 minute 02 with softdep [16:28] <@mercutio> 1 minute 50 seconds withotu softdep on vps [16:29] wow 91s on the slow FS with softdep [16:29] softdep slowed it down for you? [16:29] hmm, i have noatime too it seemms [16:30] no softdep sped up [16:31] from 800s on that guy i think [16:31] i thought you got a minute without softdep [16:31] no that was the physical host [16:31] i'm kicking a VPS at the moment [16:31] 90s on the VPS with softdep is <3 <3 <3 [16:32] testing on metal too now [16:33] what is sad is how much faster linux is :( [16:33] no kidding [16:33] like 5s on a haswell w/ssd [16:33] btrfs even :/ [16:33] i got 5.2 seconds [16:33] dman you are faster than me [16:33] lol [16:34] hey it used more than 100% cpu [16:34] tar vxzf ../src.tar.gz 2.94s user 3.04s system 114% cpu 5.222 total [16:34] 13s on a big zfs fbsd machine too [16:34] oh this was zfs on linux [16:34] ^^^^opterons [16:34] i think it's haswell too actually [16:34] yeah i7-4770 [16:34] haswell i5 here [16:35] same diff basically [16:35] i have lz4 enabled too [16:35] but i doubt it makes any significant diff [16:35] you know i saw the damnedest thing, speaking of this opteron [16:35] but yeah i wish openbsd's disk subsystem would be fixed up [16:36] the default cipher on ssh/scp is 70% the speed on the opteron as on a haswel-era pentium d [16:36] both machines capable of gig-E wire speed (verified with ioperf) [16:37] both freebsd 10.3 machines testing to the haswell linux box [16:37] cypher can make a huge diff on some machines [16:37] i don't think pentium d's have aesni [16:38] but on such capable machines that can make cpu utilisation of such vary a lot [16:38] except openbsd doesn't trust it hah [16:38] i think there's a new djb one now [16:39] chacha20 [16:39] yeah [16:39] i was about to tear that machine apart and test new NICs [16:46] anyway, thanks for the pointer mercutio [16:46] you're welcome [16:46] the stuff i've been googling for has been virt* specific [16:46] i didn't look for more general tunings because i was OK with the base-case suckage [16:46] yeah i don't think it's acutally a virtio issue [16:46] garry would be happy to hear that [16:46] openbsd is just quite far behind in disk performance [16:47] since i opened a ticket on it [16:47] there used to be a huge deficiency with smp too [16:47] that's getting better [16:47] $ doas /usr/bin/time tar xzpf src.tar.gz 501.06 real 3.78 user 11.95 sys [16:47] like the global lock was around basiclaly everything, where smp was only userland [16:47] it's getting a little finer now [16:47] <3 openbsd [16:47] i'm ok with what it is [16:48] but some times it's a head scratcher [16:48] heh [16:48] ^ so that was with a remount of /, which didn't appear to honor the -o softdep [16:48] the time seems to work but mount doesn't seem to update [16:49] ditto the physical box too [16:49] you could edit /etc/fstab and reboot and check it's still good [16:49] right [16:49] i shall [16:49] can't reboot that physical box till after hours tho [16:49] you have an arp vps right? [16:49] yes [16:49] a few [16:50] was it migrated/recent? [16:50] this was me testing the new thunder setup [16:50] very recent [16:50] ahh so that's on arp thunder [16:50] yes [16:51] i'm ok with crap performance on onesy-twosy small VPSes [16:51] i was actually considering freebsd on the thunder instance if i couldn't make it behave better [16:51] :/ [16:51] and 90 seconds on softdep? [16:52] it's running now (after reboot) [16:52] 6.0-current even [16:52] cool [16:52] 6.1 seems to be delayed. [16:52] but i don't think it'll improve performance [16:52] well i have just upgraded from 5.7 on everything, so i wasn't current on what's what [16:53] $ doas /usr/bin/time tar xzpf src.tar.gz [16:53] 115.48 rel 2.74 user 5.91 sys [16:53] daaaaaaang [16:53] i'll take it [16:53] if i'm untarring with -p, does it modify atime too? [16:53] maybe [16:53] i didn't use p [16:54] that's smoking fast considering where i came from [16:54] thanks! [16:54] and i have noatime on [16:54] heh [16:54] it's not too bad with cvs updates and the like tbh [16:54] like openbsd's disk performance used to bug me a bit on synthetic things like that [16:55] but for most things that matter like waiting for a cvs update it doesn't seem that bad [16:55] also openbsd kernel compiles are much faster than linux kernel compiels :) [16:56] I dunno, I compiled a Linux kernel (defconfig) on an SSD with -j12 in 19 seconds... [16:56] Hard to beat that. [16:56] i think openbsd is less than 2 minutes for default with one core [16:57] HAS_A_BANANA: holy cow, with what? [16:57] (CPU) [16:58] i admit i had an amd thunderbird last time i compiled a kernel for fun... [16:58] I don't really remember, it was a couple years ago... I think an i5-2400. [16:58] are you sure it didn't fail [16:58] lol [16:58] with -j12 you might not see the failure so easily [16:58] i'm compiling openbsd kernel right now :) [16:58] Yes I was sure, mercutio [16:59] make defconf; make bzImage should pretty much always work assuming your env is sane, right? [16:59] well you need as86 etc [17:00] well i feel better now. it's champagne friday. thanks all! [17:00] but yeah you'd think it'd be a tested option [17:00] (even commented on my arpnetworks ticket to that effect) [17:02] damn openbsd kernel compile is slower than it used to be [17:02] make 213.31s user 35.44s system 96% cpu 4:17.53 total [17:02] trying with make -j 3 now [17:03] actually i think gcc has been getting slower over time [17:05] make -j3 211.87s user 39.95s system 186% cpu 2:15.26 total [17:24] openbsd uses quite old gcc though [17:25] it's recent compared to 2.95? [17:25] I guess :) [17:27] 4.2.1 - only 10 years old [17:27] oh wow time flies [17:27] how old is 2.95 then? :) [17:27] 16 years [17:27] wow [17:27] 1999 [17:27] oh hangon, july 31 1999 for first release [17:30] it looks like it was 10 years ago they abanonded gcc 2.95 [17:33] looks like openbsd 5.5 was the first release to not include gcc 2.95.3 though [17:33] may 1 2014 [18:59] *** Nahual has quit IRC (Ping timeout: 258 seconds)