[00:27] kirkland: ping? [01:15] If you've got a wildcard entry in your Bind server but you want your domain without a host (say google.com instead of www.google.com) to have an entry that doesn't redirect to the same place as the wildcard, how would you go about doing that? === ujjain is now known as ujjain|afk [02:37] anyone know where i can find help with an ms exchange server? [02:38] s0rserer: ##windows? [02:38] Or your support vendor, of course [02:39] thanks ill give that a shot [02:41] evolution to exchange? [03:13] hey, how do I go back to the previous grub behaviour? - from time to time grub detects a boot failure and shows the grub menu - I don't want this to ever happen as it involves driving 40 miles to plug in a keyboard and press enter :/ [03:22] hackeron: dunno; I use hpa's bootloader [03:24] twb: what's that? [03:25] isolinux [03:25] The same bootloader you already use on CDs and USB keys, works on hard disks, too [03:25] what are the advantages to grub? [03:26] It's not a huge crufty monstrosity of unnecessary complexity [03:26] does it support software raid? [03:26] For example, it doesn't autodetect "modules" to compile into the MBR [03:26] probably won't work with software raid then [03:26] It supports linux md RAID 1. Not sure about 5. [03:26] I would not put /boot on RAID5 anyway [03:26] hmm [03:27] In fact the last straw for me with grub was that it *didn't* work properly with RAID1 [03:27] :/ - I use it with md raid on all my ubuntu server boxes [03:27] some raid1 some raid5 [03:27] Some motherboards will detect hda is dead and "rename" hdb to hda, so grub's device.list breaks and you can't boot [03:27] huh? - grub's device.list uses UUIDs [03:28] And for me, it wasn't a 40 mile drive -- the machines were in mines in .za [03:28] hackeron: maybe it uses UUIDs now, it didn't in 8.04 [03:28] Actually, those boxes would've been RHEL4 [03:28] Er, RHEL5 [03:28] In any case, syslinux doesn't have that problem, because it boots off the disk that's there, end of story. [03:29] well, this is just ubuntu's latest "feature" - far as I can tell the problem is ubuntu no longer has a time out to press esc for the grub menu - instead if instantly boots and if it fails, it marks a failboot flag and shows the grub menu on next boot - not what you want on a server :/ [03:29] Has anyone had issues getting user_dir and apache working when homes are mounted from another server? [03:29] hackeron: instead of a timeout, you have to hit Shift at exactly the time between when the BIOS hands control to grub, and grub hands control to the kernel [03:29] hackeron: which IME is impossible on some fast systems :-/ [03:30] twb: well, that wouldn't be a problem if it didn't force the grub menu on boot failure [03:30] syslinux also fixes that -- you can use Scroll Lock instead, which stays "stuck" down, so you can hit it while still in the BIOS [03:30] Sakara: mounted how, NFSv3? [03:30] twb: NFSv4 [03:30] twb: how do I "switch" to syslinux? [03:30] Sakara: I haven't tried that, sorry [03:31] http://paste.debian.net/119165/ [03:31] hackeron: those are some notes I made for two-disk RAID1 systems back in the hardy days [03:31] twb: Hopefully someone else has done it [03:32] hackeron: unfortunately Ubuntu isn't interested in making it an install-time option, though the Debian people are working on that I believe. [03:32] Sakara: could your problem simply be root_squash? Oh, are you using kerberized NFS? [03:32] yes [03:32] twb: that is what is happening [03:33] Sakara: ah, OK, so does the apache process have the necessary krb tickets? [03:33] I am unsure about how to authenticate and authorise the www-data user [03:33] Sakara: yeah, that kind of headache is why I'm not using krb yet :-) [03:33] Sakara: you might want to ask the #kerberos channel as well [03:33] what other ways are there to centralise user accounts? [03:34] I use LDAP [03:34] ok LDAP only? [03:34] And SSHFS for end users's network filesystems. [03:35] NFS is there, but only on systems where I strictly control who has root, because obviously without krb, root on every system is fully trusted by the NFS server [03:35] ok that does seem like a more appriopriate way to handle users too [03:35] On a hardy server how do I implement SMTP rate control so that the number of outgoing messages to an smtp relay do not exceed 1000 per hour. I don't want to drop the excess messages on the floor or bounce them, I just want them to be queued and delivered in the next and subsequent hours. I'm using postfix to relay the mail. [03:35] ahh ok yes tha twould be a problem [03:35] Consider: even with root_squash, root can simply "su - twb -c 'cat /home/twb/.netrc'" or so [03:36] owh: I don't know; you could also ask #postfix. I would probably run "postconf" and look for likely-looking options, then look them up in the reference docs. [03:36] <|rt|> I've always been under the impression that kerbose doesn't really centralize users...it just provides a secure authentication framework [03:37] |rt|: it can do both [03:38] twb, I've been doing that for hours now. #postfix suggest using policyd which appears to be postfix-policyd, but the documentation I can find is for v2.x, not the one that comes with hardy, v1.x [03:40] owh: postfix-policyd is 1.82-* in everything from hardy to oneiric ... are you sure you need 2.x of it? [03:41] jmarsden: I'm happy to stay with 1.82-*, but the documentation on www.policyd.org is for v2.* -- I cannot find docs/howtos for 1.82-* [03:41] Unless I missed something somewhere? [03:44] /usr/share/doc/postfix-policyd/README.gz [03:46] owh: The docs that come in the postfix-policyd package should be sufficient... I think? [03:47] jmarsden: I've read that document several times - but I might have missed something - I'm needing to limit the number of messages to a relayhost, not by sender, not by recipient, not by domain, etc. [03:48] Can't you just do the 'netblock' example, and make the 'netblock' match string be '%' so it matches all netblocks? [03:49] I have not tried it, but it looks like it ought to work for you. [03:51] jmarsden: That sounds interesting. I'm also trying to track down using transport_destination_rate_delay, which appears to work within postfix, but it's unclear how it's supposed to work, since `transport` needs to be changed to the name of your transport, which would be smtp, but I cannot find how my relayhost is linked to smtp. [03:52] Hmm, unless this line in master.cf is the culprit: [03:53] relay unix - - n - - smtp [03:56] Worst case, set both smtp_destination_rate_delay and relay_destination_rate_delay and see which one works :) [03:57] ROTFL [03:57] Thank you, I needed that. [03:58] Let me remove postfix-policyd and see if that keeps my email still working :) [04:16] twb: Isssue with authorization and the www-data uid was resovled by creating an associated kerberos principal [04:16] Sakara: nice [04:16] ty for your input it really help crack that one [04:16] Sakara: I thought it would be harder than just an addprin [04:17] I'd overlooked adding it because its not a real uid I think about alot becuase its only really used on the one machine (the web server) [04:17] cheers again [04:18] np [04:29] I am following the doc on mail server setup at https://help.ubuntu.com/community/PostfixBasicSetupHowto i can recieve mail but I can't send any ideas? [04:58] My server is doing the thing where landscape has its information repeated twice, and the second copy refuses to update or go away no matter how out of date it gets. [04:58] Which file did I need to delete to force it to refresh the information? [05:03] trimeta: landscape as in canonical's proprietary puppet analogue? [05:06] anybody used kerberos/ldap cleints on ubuntu? [05:06] New bug: #781283 in mysql-5.1 (main) "mysqlreport crashed with SIGSEGV in malloc_consolidate()" [Medium,New] https://launchpad.net/bugs/781283 [05:06] !anyone [05:06] A large amount of the first questions asked in this channel start with "Does anyone/anybody..." Why not ask your next question (the real one) and find out? See also !details, !gq, and !poll. [05:07] ok so where does kerberos failures get dumped? [05:07] messages? [05:07] cd /var/log/ [05:07] oops [05:07] hehehe [05:08] ls -l [05:08] twb: Landscape as in the thing that collects system information into /etc/motd. [05:09] Is there another name for that? [05:11] trimeta: er, no, it isn't [05:11] jmarsden: FYI, I've added smtp_destination_rate_delay to main.cf and reloaded postfix and it appears to be working. What I don't know is what side-effects this might have, but at least I've gotten one step closer -- thanks! [05:11] trimeta: try :>/etc/motd and then log in again [05:11] owh: You're welcome :) [05:11] trimeta: that's how I fixed that specific symptom on my lucid boxes (which definitely DON'T have landscape) [05:12] Does that command replace /etc/motd with an empty file? [05:12] Yes [05:12] Back it up first if you don't trust me [05:13] Right now /etc/motd is a symlink to /var/run/motd, which (according to the manpages) is created by runpart'ing /etc/update-motd.d/ [05:13] I tried moving /var/run/motd to /var/run/motd.bad, but a new /var/run/motd was created identical to the first (with the same badness). [05:13] Hum [05:14] Sorry, brain fart [05:14] I meant :>/etc/motd.tail [05:16] OK, that fixed it. Why does the spurious tail file get created? [05:19] does anybody know how to fix this I am following the doc on mail server setup at https://help.ubuntu.com/community/PostfixBasicSetupHowto i can recieve mail but I can't send any ideas? [08:05] New bug: #604593 in libpam-ldap (main) "pam_unix "account" returns success on a user with an invalid shadow password." [Undecided,Confirmed] https://launchpad.net/bugs/604593 [08:16] hey guys I have a database in cirilyc with encoding latin1. On my ubuntu machine it looks ok, but on my other ubuntu machine is all messed up. can anyone help me? [08:47] * negronjl is calling it a night === koolhead11|afk is now known as koolhead11 [09:53] hey guys, can you help me change the default charset of mysql server [10:21] New bug: #794443 in amavisd-new (main) "package amavisd-new-postfix 1:2.6.5-0ubuntu2 failed to install/upgrade: el subproceso instalado el script post-installation devolvió el código de salida de error 1" [Undecided,New] https://launchpad.net/bugs/794443 === cypha`` is now known as cypha [11:15] Is it the update-motd package which updates the message saying X number of packages need updating and System needs reboot and all that? [11:21] nigelb: try :>/etc/motd.tail [11:22] twb: I found that file, but I'm trying to figure out what updates it [11:24] I mean run the command ":>/etc/motd.tail" and the issue where you get two copies of motd will disappear [11:25] twb: oh, ah! [11:26] twb: hrm, I still have one file, but that's not up-to-date [11:29] Damn [11:29] That worked for me [11:29] Try logging in again [11:32] hi [11:32] i belive that this is not the right channel but anyway i need helt with an easy htaccess question === dfgdfg is now known as xilentares [11:34] could somebody help me ? it hast to do with url rewriting [11:34] or recommend me another channel ? [11:35] try #apache [11:36] i dont know why but i cant join it ive already tried it [12:11] come on guys nobody ? [12:14] You will never find answers to questions you never ask. [12:14] true dat [12:16] hmm okay.i have a gallery that generates a link like that index.php?id=gallery&album=test know i want that the output is /gallery/test for the user. the rest of the page is already seo optimized [12:17] my idea was : RewriteRule ^/(a-zA-Z0-9)/(a-zA-Z0-9)$ /index.php?id=$1&album=$2 [L] [12:17] but this dosnt worked :S [12:17] is that because the url is generated internal ? [12:18] hi all! I've got a strange problem with DHCP and DNS - DHCP server (ubuntu server) doesn't send DNS addresses to windows workstations [12:19] there is option domain-name-servers line in dhcpd.conf, but no workstations received DNS IPs [12:22] connection works [12:23] problems with bind9? [12:25] xilentares: Your regexes are wrong. [12:26] xilentares: You'll only ever match this url: "/a-zA-Z0-9/a-zA-Z0-9" [12:27] Good Day! [12:28] xilentares: You want something like: RewriteRule ^/([a-zA-Z0-9]+)/([a-zA-Z0-9]+)$ /index.php?id=$1&album=$2 [L] [12:34] hmm thanks soren but it dosnt work for me :( i will experiment a bit more with it^^ [12:35] xilentares: It works for me. [12:35] I just tested it. [12:35] The requested URL /gallery/test/ was not found on this server. [12:37] if i visit localhost/gallery/test it opens internal localhost/index.php?id=gallery?album=test ,right ? [12:40] New bug: #636480 in postfix (main) "Postfix and PowerDNS cannot bind to IPv6-IPs on boot" [Undecided,New] https://launchpad.net/bugs/636480 [12:42] soren : thanks got in know :) [12:43] hi! I'm running a server with ubuntu Lucid (server edition). On this server, there is a VM i use for testing and development. Now i wanted to simulate a loss of network connection for a program, i am developing and I closed the tap device of the guest VM on the host server (ifdown tap02). Now i can not bring it up again. ifup tap02 tells me "TUNSETIFF: Device or resource busy". Any suggestions? [12:45] i think there was a tcp connection open when i teared down the interface. maybe that tcp connection keeps the device busy, although its said to be down. Is there a way to flush and close this connection manually? Rebooting is not an option. [13:00] ok. got it: you can not bring up the connection, when the VM behind it is up. The VM needs to be down/stopped - then you can bring up the connection again. [13:02] good morning internets === ujjain|afk is now known as ujjain [13:52] if ubuntu domU can't see the newly assigned memory (e.g. free -m), what's the point having the ability to do it in dom0? [13:52] i am using Ubuntu 10.04, and the ballon=y in the kernel [13:52] even after i assigned more RAM to domU ubuntu, the free -m can't see it, so i am using it won't use it [14:27] Daviey: http://people.canonical.com/~chucks/gpxe-cobbler.patch [14:31] cmagina: after pushing multipath-tools to oneiric last night, it occurred to methat the 'add|change' rule in kpartx.udev is idiotic - it just modprobes dm-multipath. Which doesn't exist. [14:31] think i'll open a bug to remove all dm-multipath references, which should amount to changes to 3 files [14:34] hallyn: sounds like a good idea. i haven't looked at all those bits, but if you want another set of eyes on any of this, just toss it my way [14:34] cmagina: cool, thanks. more eyes++ [14:35] hello there [14:35] how to perform a complete hardware check on a server ? [14:38] flashlight and paper pad? [14:39] ? [14:41] tyreza: What exactly were you hoping to accomplish? [14:42] i simply want to make a complete hardware check [14:43] I don't know what that means. [14:43] What are you checking? [14:45] my system reboot automatically [14:45] when i m working on it [14:45] so i need to make a complete check [14:46] so my question how ? [15:07] ? anyone there [15:08] i can't see the difficulty on my question ? [15:08] hi! [15:08] it is easier to speak directly to a person [15:11] guillemhs: just ask your question, if someone knows, they'll respond. [15:16] Hi [15:16] hi [15:17] pmatulis: Any idea about the availability of JAVA6 U26 ? [15:17] on Ubuntu Server 10.04 LTS [15:17] kaushal: i'm sorry, i don't know what JAVA6 U26 is [15:18] ok [15:19] cmagina: nm, i guess we should support custom kernels :) leaving that alone [15:21] hallyn: ah, yeah, didn't think about that [15:21] hallyn: good catch [15:27] New bug: #656421 in samba (main) "No DNS Updates on Domain-Registered Server in Win2k8R2 enviroment" [Undecided,New] https://launchpad.net/bugs/656421 [15:43] hello, how can I show what partition / is on? it doesn't show up in df or fstab. [15:50] cthompson: mount ? [15:55] well, it looks like I had to edit fstab to allow the kernel to mount / on a raid partition /dev/md0, after that: df: /dev/md0 on / (all good) [15:56] thanks though [15:59] * genii-around sips his coffee and ponders if cthompson has raid modules in his initrd [16:53] ho. [16:59] Daviey: hu [16:59] RoAkSoAx: o/ [16:59] Daviey: o/ how's it going today man? [17:02] RoAkSoAx: not so good.. primary laptop power cable busted. [17:02] Yourself sir? === koolhead11 is now known as koolhead11|afk [17:12] Daviey: darn :/ [17:13] doo-do-be-do [17:13] afternoon [17:13] TREllis, RoAkSoAx, zul, negronjl [17:13] talk is about squid-deb-proxy + cobbler on orchestra [17:13] orchestra has some cobbler snippets and a mini.iso importer for new distros [17:13] RoAkSoAx: you were saying? :) [17:13] zul: you pointed it needed to work with debian and d-i instead of kickstart, right? [17:13] hey Daviey o/ [17:15] right because thats what we use [17:15] cobbler uses debmirror to do repository mirroring, at the moment we use deb-squid-proxy as a cache instead [17:16] TREllis: debmirror works now [17:16] * zul lunches [17:16] TREllis: the only thing, is you want to install anything using that local mirror [17:16] zul: bon appetit! [17:16] s/is/if [17:16] you need to tell the preseed that the mirror is there [17:16] RoAkSoAx: yeah you'd need to tweak the preseed [17:16] TREllis: yeah, so using snippets we can do that automatically, but we need to find a way of doing so :) [17:17] TREllis: that's why I was saying, whenever there's a local mirror for a profile, use that snippet/preseed/wathever [17:17] RoAkSoAx: so basically, we need a snippet that checks if a mirror is present and uses it if it is, otherwise use the proxy [17:17] TREllis: if there's not, and we now that we are using squid-deb-proxy, use the preseed for that purpose [17:17] TREllis: correct [17:17] my working with cheetah is quite limited, you can probably do it though [17:17] RoAkSoAx: you want to take that as an action item? [17:18] got a weird problem with ubuntu 11 and vmware fusion. after upgrading to 11, my terminal does not work at all. i can still ssh in, but the local terminal is just blank. [17:18] no X. the VM has been upgraded through a few versions of ubuntu. [17:18] oh, wait, there's simply a blinking cursor up in the top left now. sometimes it's there, sometimes it's not. [17:19] RoAkSoAx: where are we keeping track of the actions items for this btw :> [17:19] ok, now i see that it's simply defaulting to tty7, which is blank. [17:21] lynxman: sure, I'm currently finishing the repository thing when adding an ubuntu mini.iso.... and checking the debmirror stuff is working well [17:21] RoAkSoAx: excellent :) [17:21] then I'll document the preseed changes needed to use the local mirror [17:21] lynxman: and we can get the snippets after that [17:21] sounds good [17:22] RoAkSoAx: sounds good, you're the best... arooound [17:22] RoAkSoAx: sorry, grew up in the 80ies [17:22] other than that, mini.iso import detection should be done automatically this cycle, as cjwatson agreed on providing us with something to allow detection [17:22] so we can drop having to manually specify breed, os version, etc [17:23] RoAkSoAx: that would mean submitting a patch for cobbler, the way it's detected now is through the directory structure [17:23] RoAkSoAx: it's easy to patch, I can do that even :) [17:23] lynxman: yeah I'll take care of that too [17:23] RoAkSoAx: rocking [17:23] lynxman: the idea is to submit to upstream most of the stuff we do here [17:23] off course, not the orchestra part but cobbler itself [17:23] RoAkSoAx: that's always the idea :) [17:24] lynxman: what else is worrying you? [17:24] :) [17:25] RoAkSoAx: we decided at UDS that we would use a bit more the cobbler API in order to store somewhere hardware profiles for the hardware in the orchestra server network [17:26] lynxman: +1 .. have you guys had a chance to see if thats even feasible ? [17:26] RoAkSoAx: would like to get your initial ideas about how to attack that, I know this is a long term project thinking, but wanted to know your ideas [17:26] lynxman: i believe that would be using the "System" concept in cobbler [17:26] SpamapS: not yet, we're knee deep into making all the parts work well for our demo at Structure in 2 weeks :) [17:26] indeed [17:27] lynxman: have you guys looked into the "Systems" [17:27] RoAkSoAx: not yet, that's why I wanted your initial ideas, so I know where to look in ;) [17:28] lynxman: http://docs.redhat.com/docs/en-US/Red_Hat_Network_Satellite/5.3/html/Reference_Guide/s1-cobbler-addsystem.html [17:28] provides a brief, understandable introduction to it :) [17:28] RoAkSoAx: thanks :) [17:28] *bookmarks it* [17:28] lynxman: system profiles in general are cool, as long as you have the mac address of a system, you can assign a kickstart/preseed for it to use [17:29] TREllis: you cannot use cobbler deploy though to deploy a machine AFAIK [17:29] hey lynxman o/ [17:29] TREllis: the idea was to have a specialized mini boot image that would just boot up, compile system information and report back then shut down [17:29] RoAkSoAx: yeah, uses the fence tools right? [17:29] lynxman: that's a neat idea, I've seen a similar thing used in a bank before [17:29] oh btw.. fence-agents have been already uploaded and are in the NEW queue [17:29] so we should see them next week [17:29] then once we got the info we can just install the basic image and from there take a logical decision towards what kind of application can be fit in the hardware [17:29] TREllis: yes and no... the "deploy" feature has been disabled upstream [17:30] TREllis: "This feature was removed from Cobbler 2.0 and may be present in a future release. It is not yet implemented in the code as I wanted to revisit and rewrite much of the core implementation -- Michael" [17:30] lynxman: is there going to be somewhere to store that data in orchestra then? [17:30] in https://fedorahosted.org/cobbler/wiki/DeployFeature [17:30] RoAkSoAx: fair enough [17:30] TREllis: that's another different dimension :) [17:30] TREllis: I reckon we'll store it in mysql or such, but at some point we want that to evolve into the metadata service [17:31] lynxman: if the debmirror local mirror is presented over http... squid-deb=proxy could use that as it's upstream mirror. Perhaps wasteful.. but easier to implement, and the autodiscovery magic would still work. [17:31] RoAkSoAx: and since that was michael, we can assume it's been disabled for a while then and hasn't been looked at [17:31] TREllis: indeed! [17:31] Daviey: very good point! [17:31] Daviey: duplicate data though? [17:31] Daviey lynxman yeah but that would be using snippets to tweak it anyway [17:31] Daviey: ie you need twice the storage, once for the mirror once for the cache [17:32] lynxman: I have a python script for adding new hardware to cobbler btw [17:32] i need to add facter output to it. [17:32] Daviey: oooh, can we see it somewhere? [17:32] TREllis: yeah [17:32] Daviey: if we have a local mirror, then ew don't really need squid-deb-proxy [17:33] RoAkSoAx: well, unless you are doing deployments over long distance links, mirror in one DC, cache in the other [17:33] RoAkSoAx: what we like about squid-deb-proxy is proxy auto discovery [17:33] lynxman: yeah [17:33] lynxman: it's only simple atm, i was using it to experiement with the RPC api. Seems to work well. It's for the auto discovery WI [17:33] TREllis: yeah but AFAIK, for that reason you;d have a second cobbler server/mirror [17:34] TREllis: that's the whole point of having a local mirror :) [17:34] not use long distance links :) [17:34] RoAkSoAx: you'd only get the hit on the cache once :) [17:34] I know plenty of people who do that [17:34] TREllis: in that example, the squid-deb-proxy cache size could be tweaked to be very minimal.. perhaps 0. However, TBH, for a base install - we are not talking about that much TBH. [17:35] Daviey: yeah, very true [17:35] calling mvo into this, maybe he has some good ideas [17:35] I think we should have a call about this soon.. [17:35] ideally this week. [17:35] who wants to be on it? [17:36] TREllis: right, but in that case we have a server for solely be a mirror cache [17:36] Daviey: me :) [17:36] RoAkSoAx: true. I like the idea of the snippet determining if you have a mirror or cache setup [17:37] TREllis: yeah [17:37] TREllis: indeed, it should be smart enough [17:38] now my main concern is the cobbler mirroring + squid-deb-proxy integration [17:38] or if we should just scratch squid-deb-proxy from orchestra and go straight for cobbler mirroring [17:38] If cobbler is a local mirror, why do you need s-d-p ? [17:38] Daviey: would love to listen in, but heading to the IoW festival :-) [17:38] s-d-p is for when you don't want a local mirror [17:38] SpamapS: exactly what I was asking :) [17:39] TREllis: that's what you cool kids call it now? :D [17:39] TREllis: slack. [17:40] Daviey: he totally is [17:40] /part [17:40] :-) [17:40] TREllis: you had one space in front of that part [17:40] /kick TREllis too slack for here. [17:40] SpamapS: agreed, guess that leaves it up to the snippet then: if debmirror is installed, point to it; else use s-d-p [17:41] TREllis: yeah but s-d-p is installed by default on the client side, how we can take a smartz decision like that from the package? [17:41] Daviey: it's a hard live, but someone has to do it [17:41] I'd like to hear mvo's view.. but i'd like to experiement with have s-d-p as the primary interface.. With either a local mirror (debproxy) or primary external mirror providing it. [17:41] TREllis: indeed. [17:42] s/debproxy/debmirror ? [17:42] since that's what cobbler uses by default [17:42] Daviey: messaged mvo, I think it's doable... [17:42] lynxman: can you ask him to join here, so we can all jump on him? [17:42] :) [17:42] Daviey: right, but using s-d-p when we have a local mirror doesn't really make sense [17:42] Daviey: I did that 10 mins ago, I guess he's afraid... or having dinner. I'll go with afraid [17:43] are we pulling in debmirror 2.7 to oneiric? has a kickass feature --debmarshal [17:43] cause you'd be cashing, what we already have "cached" locally [17:43] Why would I want to go ->squid->debmirror ? [17:43] RoAkSoAx: no.. it doesn't make lots of sense.. but if it's our primary interface it should be investigated IMO. [17:44] SpamapS: it's just that we install s-d-p-client as default on the orchestra client, we were seeing a smart way to be able to use either debmirror or s-d-p as required [17:44] TREllis: don't complicate things with asking for debmarshal support :) [17:44] Daviey: we want debmarshal support *hides* [17:44] I agree with the keeping it simple part. [17:44] negronjl: +1 [17:44] Daviey: hehe, <3 debmarshal [17:45] Grr. [17:45] Daviey: although I moved my local mirror to reprepro anyway [17:45] lynxman: about the hardware profiles cobbler has couchdb support but its not secure at all ust yet [17:45] TREllis: sane here :) [17:45] Ahhhhhhhhhhhhhhhhhhhhhhhhhhhhh! [17:45] zul: maybe it would just be easier to have a mysql plugin? it's extra easy [17:45] TREllis: stop scaring Daviey please [17:45] lynxman: ergh... [17:46] couchdb keeps me awake at night. [17:46] Since cobbler is going to seed the thing.. s-d-p client is actually, IMO, kind of unnecessary. [17:46] zul: I know you like it, but couchdb... bleh [17:46] zul, lynxman: mysql +1 [17:46] lynxman: sqlite database would be easier [17:46] i swear there is a couchdb hiding under my bed. [17:46] The advantage of s-d-p is just that its specifically only allowing proxying of ubuntu. [17:46] zul: don't get me started on sqlite [17:46] Daviey: but that's the thing, whenever someone doesn't want to have the local mirror in full... there's no need to do so and we can use s-d-p for that [17:46] SpamapS: exactly [17:46] adding couchdb would add more complexity plus pretty much everyone already knows mysql [17:46] Daviey: but if someone *wants* to have their own local mirror [17:46] zul: just trying to think 2 steps forward towards the metadata service [17:46] Daviey: then they should be able to without having the hasle of installing s-d-p client [17:47] And why again aren't we just using the data storage capabilities built into cobbler? [17:47] thats what i said [17:47] RoAkSoAx: it's installed by default via the client package. [17:48] SpamapS: have you used that? [17:48] What is the *main* reason for having s-d-p/debmirror in orchestra ? [17:48] negronjl: not having n-hundred machines pulling packages [17:48] Daviey: I know :) but still if someone wants to sync the whole mirror for one release, they should be able to do so [17:48] negronjl: so reduce bandwidth consumption drastically [17:48] ok....so, find the simplest way to accomplish *that* [17:48] Daviey: but if they do so and use it, s-d-p client has no point [17:48] it seems that the conversation is going in multiple directions at the same time :/ [17:48] negronjl: *g* [17:49] RoAkSoAx: ack, i'm not disagreeing with that [17:49] Daviey: IIRC, it is just json files on disk. But the point is that its already got a database of machines that I am provisioning. To have two databases of machines I'm provisioning, instead of enhancing the first.. seems a bit failure prone. [17:49] negronjl: it is [17:49] Daviey: yeah that';s why we should keep both as options and make sure they both work well :) [17:49] SpamapS: hmm. i'm not sure i follow. [17:49] I agree with negronjl [17:49] SpamapS: Daviey: So how about we see the different approach scenarios for the hardware database and we all decide which one is the one we should proceed forward with? [17:49] two conversations are intermingled [17:49] a) proxying/mirroring [17:50] b) metadata [17:50] lets just wrap up the a) before talking about b [17:50] c) hardware database and then c [17:51] zul: thats the same thing as metadata [17:51] Okay.. I am proposing that we investigate always using squid-deb=proxy. The parent mirror can either be a local archive, or a public shared one. [17:51] Yes, there is potentially a wasteful extra layer in the stack. [17:51] However it should make things simplier to construct. [17:52] Daviey: could this be useful you think in order to let the user decide if they want s-d-p or a full mirror? [17:52] lynxman: Well i am proposing that we always using s-d-p, regardless of what feeds it. [17:52]