[00:17] why know what is DNS server? I've read about BIND if I install it what I'll get? [00:18] Headaches [00:18] Neo4: there's three types of DNS servers: authoritative, recursive, and forwarding [00:18] genii: lol +1 [00:18] what does do DNS server? it return IP of servers [00:19] Neo4: an authoritative server knows the IPs and names of specific services and so on [00:19] Neo4: Basically, yes. [00:19] Neo4: a recursive server knows how to start from the "root nameservers" and query each hierarchy of name servers to find an eventual answer to a question [00:19] sarnold: ok [00:19] Neo4: forwarding nameservers do not know how to query the roots, they just forward the question on "to the next nameserver", which *will* know how to answer the question [00:19] in general what shall I get? [00:20] Neo4: so which types of DNS server you need to run depend upon what services you want to offer [00:20] Neo4: if you have clients on a LAN that want to look up hostnames like www.yandex.ru, then you would probably want to run a recursive or a forwarding server for your clients [00:20] I will able put any domains to my DNS server? something like ns.my_fqdn [00:21] Neo4: if you want to provide services to the world with your own names, you could run an authoritative server [00:21] sarnold: I needn't any ) [00:21] yes [00:21] just curious [00:21] if you want to run your own name servers, you would register your name servers with your registrar ("glue records") [00:22] Neo4: I strongly recommend powerdns instead of bind [00:22] Neo4: other popular choices are knot and unbound [00:22] sarnold: I want for test run one on digital ocean and overview all possibilities and settings ) [00:23] Neo4: if you set up a recursive server DO NOT MAKE IT PUBLIC [00:23] on digital ocean popular BIND [00:23] bind was the first and still very popular [00:23] but they mix auth and recursive which has shown to be very dangerous [00:24] i've read both bind and powerdns sources and I know which one I'd rather run :) [00:24] unbound is fairly decent [00:24] I concur, unbound is an excellent recursor [00:25] I have yet to try powerdns but I think sarnold just convinced me ;) [00:26] if I have my own DNS somewhere I can bind there domainname and ip address from digitalocean and not use a cname and other from digitalocean panel? [00:26] powerdns folks also make a dns proxy, dnsdist -- during a recent round of dns DDOS attacks, folks with bind servers were able to servive by plopping dnsdist services in front [00:26] it might be this functionality is DNS [00:27] badly understand this notions [00:27] Neo4: you need to get IPs and IP routing from somewhere.. [00:32] * mason is a staunch BIND fan. Goes with the whole dinosaur thing. BIND and Sendmail forever! [00:32] mason: let me guess, *real* sendmail, no m4 for you? :) [00:32] No, I'm an M4 fan. In fact, I've had two customer issues come up recently where I got to spread the Sendmail love. [00:33] hehe, that's greatdnl [00:35] hah [00:39] rbasak: hrm, did you not see this? https://paste.ubuntu.com/26455183/ [00:39] rbasak: makes the gpgv stuff ... unclear how to use. The manpage implies 2 is a fatal error [00:43] On 16.04 the man page for unattended-upgrade says /etc/cron.daily/apt initiates the upgrade process. Anyone know why /etc/cron.daily/apt isn't generating for me? Has anyone else seen this? [00:53] nacc: I didn't see that in my testing. Perhaps you're using an older series than I di? [00:53] nacc: if so it's still a valid problem though. [00:54] Might be able to work around by providing the DSA public key too [00:55] hashwagon: looks like it's a systemd timer thing now [00:55] hashwagon: check out systemctl cat apt-daily-upgrade.service [01:13] sarnold: What's wrong with real sendmail? It's simple and efficient! [01:15] genii: "simple"? :) this is the first I've ever heard that word used with sendmail :) [01:15] Monolithic, single binary, single process. Few moving parts. [01:15] That it's self-aware is incidental. [01:17] hehe [01:49] sendmail.cf contains enough moving parts to make up for that. [04:16] hey y'all, I want to set up bind or dnsmasq or some such thing [04:16] and I want it to act as a dns cache [04:17] which just looks up non-cached or out of TTL type thing [04:20] MJCD: I like unbound for that sort of role. [04:23] rbasak: i was checking xenial-updates as a random test on bionic [04:24] rbasak: is the DSA public key available via a different keyring? [04:24] mason, oooh [04:24] let me google that [04:25] ohhhh [04:25] this looks great [04:25] and its recursive [04:29] yeah mason this is exactly perfect [04:29] I can set upto 4 forward-addr [04:30] which by default is already google dns [04:30] brilliant [04:30] MJCD: Good, glad you like it! [04:30] thanks so much [04:30] see y'all soon [04:30] o/ [04:32] rbasak: oh i see what i was doing wrong, i need to pass all the keyrings [06:09] hi guys [06:09] i'm struggling with dhclient/ipv6 and wonder if i hit a bug or not [06:10] DHCP Client System: trusty server, 4.4.0-111-generic, isc-dhcp-client 4.2.4-7ubuntu12.10 [06:10] host gets ip6 addr normally: "ip a s" inet6 2003:.../128 scope global valid_lft forever preferred_lft forever [06:10] but after some time (probably has something to do with lease-time?) the ip becommed depreffered [06:10] preferred_lft is set to zero and in syslog i can see: dhclient: PRC: Address 2003:... depreferred [06:10] but this does also happen on 16.04 server [06:11] this is why i am not sure if its a bug or "working as designed" [06:11] does anybody know whats happening there? [06:12] RAs are beeing sent from the firewall and routes are refreshed normally. [06:22] Hi, I'm trying to get crontab to work. I have the following: `@reboot and inside of run, there is a bash script with `#!/bin/bash and then it cds into a dir and starts a nodejs script. But it doesnt work for some reason. If I try to do `/home/sam/Documents/repo/run` as a regular user, it works fine. [06:46] thanks nacc for the importer reset, AFAIK the missing versions were much older but I'll report next time I see something [07:21] Good morning [07:39] good morning [08:12] Hey cpaelzer [08:12] How are you doing? [08:22] hi lordievader, doing good for now [08:22] as soon as all I work on is built I'll face the wall of errors that I expect :-) [08:53] i have a LXD profile that had limits.memory.swap set to false. I've changed that to true. do i need to restart my containers for that setting to have any effect? [09:03] you can check if it directly applied via lxc config get limits.memory.swap [09:04] I pinged a few friends who should know the answer about the restart [09:04] hopefully one shows up in a bit [09:04] disposable2: ^^ [09:05] cpaelzer: thank you. i had tried the lxc config get before i asked but it doesn't return anything [09:05] only an empty line [09:05] for me as well, as soon as I set something it obviously retruns what I set [09:07] I wonder what it tweaks in cgroups - is it only per continaer swappiness? [09:07] if so that would eb easy to check [09:08] disposable2: yes that is it [09:08] what is your /sys/fs/cgroup/memory/lxc//memory.swappiness [09:09] switching this off/on seems to swicth that between 0 and 50 [09:09] so once you edited your profile from false to true, check if the value changed from 0 to 50 [09:09] disposable2: ^^ [09:10] cpaelzer: well, now that i've set it manually for all my containers, it says 50 for all the containers [09:10] which is the value for "true" [09:11] cpaelzer: yet, the ram is almost completely full and no swapping is happening. the host machine has vm.swappiness=60. hmmmmmm [09:11] that is the global default value [09:11] disposable2: and it will still swap only what it considers rather inactive [09:11] if you e.g. have cold page cache that will be dropped first [09:12] disposable2: if this is not your prod machine you can check if/when it would swap by using a mem eater keeping his memory hot and slowly increasing its size [09:13] cpaelzer: thank you for taking the time [09:13] stress-ng --vm-keep --vm 1 --vm-bytes can do that for you [09:14] I had my share of fun with swap in the past and experience sharing is part of the open source spriti right :-) [11:23] What would be the most sensible way to permanently change CPU scale governor? I found multiple ways: udev rule, sysfsutils (can't I use sysctl?) or just dump it into rc.local? [11:30] xnox: hi, about my ping yesterday about ocfs2-tools on s390x [11:32] xnox: I have an ocfs2-tools ftbfs upload stuck in excuses because the s390x tests fail. You filed https://github.com/markfasheh/ocfs2-tools/issues/22 [11:35] I mean, my upload fixes the ftbfs :) [11:35] I fixed my problem, turns out it was supposed to be on "ondemand" but system only has performance/powersave and it choose powersave as fallback [11:43] But on the topic of that: who defines these scaling governors? system or cpu? [11:46] soahccc: have you tried cpufrequtils? (Sorry, didn't get the whole context) [11:46] and/or cpufreqd [11:47] ahasenack: yea it's cpufrequtils (included in the image from the hoster) but they have ondemand in there but the new CPUs in our new servers don't have that [11:48] and cpufreqd? Can't you chose a governor there and it will set it every time it starts, i.e., at every reboot? [11:49] that being said, my artful system has this: [11:49] /lib/systemd/system/ondemand.service:ExecStart=/lib/systemd/set-cpufreq [11:49] there is no cpufreqd but I edited (and found it) in /etc/default/cpufrequtils [11:50] which runs /lib/systemd/set-cpufreq [11:50] curiously there is a service "ondemand" which I guess should set governor to ondemand, no idea if I need that service for anything now [11:50] do you have that systemd file above? [11:50] maybe debug it, because it looks like it tries to do the right thing [11:51] FIRSTCPU=`cut -f1 -d- /sys/devices/system/cpu/online` [11:51] AVAILABLE="/sys/devices/system/cpu/cpu$FIRSTCPU/cpufreq/scaling_available_governors" [11:51] check what you get for $AVAILABLE [11:51] I have: [11:51] $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors [11:51] performance powersave [11:51] in a laptop, of course [11:52] I don't have both of these files [11:52] not even the /sys/devices/system/cpu/..... one? [11:52] and sys reports only two governors [11:53] is this xenial or what? [11:53] performance and powersave (same as cpufreq-info says), xenial yes [11:53] https://gist.github.com/2called-chaos/03263073f6d3ab83a9b9f72ee4a244f1 [11:53] in xenial you have /etc/init.d/ondemand? [11:54] yes, that's there [11:54] it's similar code [11:54] that is what is setting your governor [11:55] you need it to be set to performance? [11:55] ahasenack: I assume it uses /etc/default/cpufreqinfo no? Because it was set to ondemand there and I guess it defaulted to powersave? [11:55] it does not [11:56] look at the script, it does not read /etc/default/cpufreqinfo [11:56] /etc/default/cpufreqinfo must come from another package [11:56] ahasenack: yeah and it has a comment in the file that it's from the hoster's installimage [11:57] I haven't restarted the machine yet (I set it to performance manually) but here's where I changed it to performance: https://gist.github.com/2called-chaos/457ee50f08df3a1b25059bedb80ba234 [11:57] I don't see a way in that /etc/init.d/ondemand script for it to set the governor to performance [11:57] it's either interactive, ondemand, or powersave. If your system supports neither, it exits without touching the governor [11:58] I restarted ondemand service and it didn't change back [11:58] which package profides that file? dpkg -S /etc/default/cpufrequtils [11:58] provides* [11:58] and then check if the package has an initscript or something like that, with dpkg -L [11:59] no path found matching pattern /etc/default/cpufrequtils [11:59] but I guess it's from cpufrequtils (same name) [11:59] makes sense [11:59] i cpufrequtils - utilities to deal with the cpufreq Linux kernel feature [11:59] look for an initscript in it [11:59] and then check if it reads /etc/default/cpufreqinfo [11:59] or just do grep /etc/default/cpufreqinfo /etc/init.d/* [12:00] could also be an upstart job. Then do grep /etc/default/cpufreqinfo /etc/init/* [12:00] and /lib/systemd/system/* for systemd [12:00] yeah there is and it does :) mystery solved (no idea why ondemand is there though) [12:00] then that initscript should set it [12:01] the only other possible problem is if it comes before the ondemand initscript, as the ondemand one could override the changes [12:03] I think the ondemand service is broken, the script reads AVAILABLE and DOWN_FACTOR variables, the latter doesn't exist [12:04] But do I even need that service if there is apparently a different service doing the same thing? [12:05] it's part of the initscripts package, so you can't just remove it [12:05] you can disable it [12:05] if your cpufrequtils one comes after, though, there is no harm in keeping both [12:22] ahasenack: haha these fools, they misspelled "govenor" variable, their script wouldn't do shit even if I had ondemand [12:22] which script? From ubuntu, or from your provider? [12:22] from the provider :D [12:22] heh [12:23] well, mistakes happen [12:23] I'm glad you found out :) [12:24] took us 2 weeks actually. we ordered new servers and our page got slower. we were like "okay, microcode update, PTI and slightly worse single core performance"... yesterday I imported 500 million records and the page was faster and we were like ._. [12:41] rbasak: hi, could you please (re)import gvfs into git? It's stale: bionic has 1.34.1-1ubuntu4, ubuntu/devel is at 1.32.1-0ubuntu1, and there is no bionic branch [12:46] ahasenack: running [12:46] We concluded that the importer had been stuck a while. [12:47] rbasak: when it breaks like that, it's really stuck, or crashed? [12:47] I think Nish restarted it yesterday, but that's why it's behind on so many packages. [12:47] a crashing importer is easier to handle than a stuck one [12:47] It hangs on talking to Launchpad [12:47] mh [12:47] I think Nish also filed a bug to investigate where we need to fix the timeouts [12:47] yes [12:47] I think it's within launchpadlib somewhere [12:47] https://bugs.launchpad.net/usd-importer/+bug/1745211 [12:47] Launchpad bug 1745211 in usd-importer "launchpad outages hang the importer and scripts calling into launchpadlib" [Undecided,New] [12:48] cpaelzer just pointed me at it [12:48] at least our answers are in sync [13:02] any help on this would be very appreciated -> https://unix.stackexchange.com/questions/419104/what-is-partuuid-from-blkid-when-using-msdos-partition-table/419116#419116 [13:02] please look at my comment of the answer [13:03] i'm trying to kickstart install without breaking existing win7 install [13:03] i'm not a end user, i need it as deploy tool [13:03] without any manual intervention [13:05] I have had a few minor situations where 'atftp' package dies. Normally I would just use systemd and make sure the mode is restarted, or in the past used something like monit / runit to make sure the service stays up. So in the case of atftp it has an init.d script which is absorbed by systemd and ran. I would normally replace this, but is there a way of extending the option. So I can add a parameter like restart [13:05] always ? [13:05] Like a systemd extends for example [14:04] boxrick: Is https://askubuntu.com/questions/659267/how-do-i-override-or-configure-systemd-services what you're looking for? [15:40] jamespage: i got started on b3 deps for queens. here's the list of what remains: https://paste.ubuntu.com/26459041/ . i still have a few i'm wrapping up that aren't in that list. [16:09] Nivex: hi, autofs uploaded to bionic :) [16:09] rock on! [16:10] You want another easy one? :) [16:12] sure [16:12] https://bugs.launchpad.net/ubuntu/+source/partman-iscsi/+bug/1641656 [16:12] Launchpad bug 1641656 in partman-iscsi (Ubuntu) "initramfs parameters invalid for IPv6 portal" [Undecided,New] [18:41] rbasak: sorry, i've been afk on nhouse stuff; did you want to sync today still? [18:49] nacc: I'm tied up this evening now, sorry (not you - the team meeting running over and then hit my EOD) [18:50] rbasak: totally fine; i did get one test written that ensures we are using the right URL for the Release -> Sources lookup [18:50] rbasak: i'll see if the scripts are dtrt, and i'll put up a MP for you to look at and we can discuss further tests from there === devil is now known as Guest81887 [22:45] is there a way to ask systemd to sanity check a given unit? [22:46] sdeziel: systemd-analyze verify ? [22:46] sdeziel: per https://github.com/systemd/systemd/issues/3677 [22:47] sdeziel: not sure how far you want the sanity checked :) [22:47] nacc: I'm looking for a tool that will sanity check a unit and any override snippets it may have [22:47] sdeziel: the above will only check the syntax, afaik [22:49] nacc: indeed and it doesn't check the $foo.service.d directories either [22:49] thanks anyway, I'll keep digging the various man pages [22:49] ship it all to another system and try? [22:50] sarnold: I'm cooking a puppet module to let one drop some override snippets then trigger a service restart. The sanity check is to avoid the foot gun ;) [22:51] puppet step number one .. spin up a new server somewhere .. [22:51] hehe [22:52] https://memegenerator.net/instance/55819969/chuck-norris-meme-testing-is-for-wimps-real-men-test-in-production [22:53] :) [22:54] http://i1.wp.com/agilescout.com/wp-content/uploads/2012/05/i-dont-test-my-code.jpg [22:56] wow, someone loved it enough to create this https://www.idontalwaystestmycode.com/ [23:04] I didn't know there was another way to test code [23:06] "systemctl daemon-reload" will catch any typo in the unit but it's then too late and the bad file will be deployed [23:19] nacc: I submitted https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=888463 to debian [23:19] Debian bug 888463 in bind9utils "bind9utils: missing python3-ply dependency for python scripts" [Normal,Open] [23:20] tomorrow I might check what's really going on: why debhelper didn't catch that [23:23] ahasenack: what fille specifies ot upstream (e.g. requirements.txt) that ply should be used? [23:26] ahasenack: it should be generated by python3:depends, aiui [23:27] yeah, but it's not working. Not in ubuntu, nor in debian sid [23:27] something with bind's build system probably