[00:01] I don't see why [00:02] Admittedly I've never tried it with a dynamic IP on the hub. [00:16] kees: go ahead and publich eucalyptus, all is good [00:29] hggdh: thanks [01:11] Can anyone easily tell me when mountall-varrun.conf was added to lucid? [01:11] That is, /etc/init/mountall-varrun.conf [01:12] hello, I know ubuntu-cloud require's 2 server's to run, Does it require them to be in the same location? I am asking becouse I am looking at renting some rackspace from a company, and i would rather have only one server there [01:12] Hmm, changelog claims mountall 2.0 [01:20] timholum1, 3 as far as I know :) [01:20] I dunno the real names, but, a director, a storage node, and a machine node [01:21] patdk-lap: I thought I read someware basicly just a management node, and a machine node [01:22] cluster controller, storage controller, and node controllers [01:22] your missing the storage node then that actually has the harddrives [01:22] but in the example you where reading they might of loaded both of them onto the same machine [01:23] ahh, ok. hmm I might have to look at a different option then :( === jjohansen is now known as jj-afk [01:23] maybe install xen or kvm and find a web interface that can manage them [01:24] ya, looks like the smallest setup is two machines [01:24] one runs everything, except the node controller [01:24] so that means the cluster controller also has the storage controller on it [01:24] and I suppose that would not be the best performance over the net :) [01:25] depends :) [01:25] ( internet not network :) ) [01:25] oh defently not :) [01:25] ya, I am defenetly going to have to look into a solution all in one box [01:27] my current budget for the project is about $3k and the solution has to be rackmounted in less then 2U's [01:27] heh, you can get 4 systems in a 2u case [01:27] but your talking about 12-20k [01:28] I guess you might be able to find a dual system case somewhere [01:28] or just use 2 1u servers [01:28] :) ya if I had 12 - 20k I could build some awsome server's [01:29] ya, but then I am wasting about 1k on a server that is only going to manage the other one, [01:29] then your not building a cloud :) [01:30] I already have the server picked out I want :) dual 8 core amd opteron with 32G of ram :) [01:30] a cloud has several nodes, plus one or more that checks that the nodes work, and failover them as needed [01:30] well it does not have to be a cloud persay, just a machine that has the ability to run vm's [01:30] :) [01:30] well, just install kvm :) [01:30] Citrix Xenserver would actuly meet all of my requirements [01:30] xen is slow :( [01:31] buy ya I am leaning tword kvm as well [01:31] kvm is fast, just run through memory quick [01:31] I thought kvm was better with memory management then xen [01:31] it may be better at it, but it still can't overcommit [01:32] :( [01:32] xen has some limited overcommit support [01:32] O well over commit is not a deal braker at this point [01:32] atleast kvm has memory dedup, that helps alittle [01:33] I am mostly interested in performance [01:34] the server I want has 32G and is upgradable to 128G which is plenty for what I am going to be running [01:35] hi guys, i have a ubuntu server 10.04 installed running tomcat and postgres, and sometimes the network hang up, cannot ping, it's not acessible at all [01:35] remember, your are unlikely to ever upgrade it past half of what it is able to do [01:35] due to memory pricing of the largest density ram being ungodly expensive, and well, replaced with something else quickly [01:35] I am going to be running 1-2 email server 1-2 asterisk server and 4 -6 lamp servers [01:36] hmm, I wonder how asterisk works in a vm these days [01:36] works great [01:36] I did not like it at all, a few years ago, only raw hardware [01:36] I am using voip.ms as a phone provider [01:36] so they have the tie in with the phone system [01:37] I just set up an IAX channel to them [01:38] my virtual server ( through rackspace cloud ) version of asterisk actuly sounds better then my in house one [01:39] my in house one is running asterisk 1.4 and my cloud one is running 1.8 so that is probably why [01:39] hi guys, i have a ubuntu server 10.04 installed running tomcat and postgres, and sometimes the network hang up, cannot ping, it's not acessible at all [01:41] lucasreborn: have you tested your nic and switch? [01:41] yeah [01:41] separetely i have another fileserver, and a another firewall [01:42] it's just my webserver [01:42] i need to do networking restart every time; [01:42] hmm [01:42] sometimes appear at the console: error in task cp:"pid", task blocked for 120 seconds, or something like this [01:43] it's so randomly [01:44] what would be your test? [01:44] when it happens how much free memory do you have? [01:45] and how close to full is your disk? [01:45] my disk usage is 8%, i have a NFS server with all data, using ~5% atm [01:45] i have 8gb, most of time, have 1gb free. [01:47] i fix doing a networking restart, have something to do about memory? [01:48] I have just seen used up memory and used up disk space do some weired errors before [01:48] and unfortuanatly I have never run tomcat or postgresql. [01:49] plus 95% of my servers are centos, I just came in this room to research ubuntu-cloud [01:49] what you mean "testing nic or switch"? [01:49] i mean, how to do a good testing? [01:50] I would personaly go into /var/log and poke around in the log files, look for things around the same time that your network went donw [01:51] I would also run memtest and a hardrive test on the system, just to eliminate bad hardware ( which can cause a ton of weired issues ) [01:51] for memtest and a bunch of hardrive tests google for ultimate boot cd [01:51] [23:49] what you mean "testing nic or switch"? [23:49] i mean, how to do a good testing [01:53] the switch does not apper to be the culprate due to you having other systems on that same switch that do not go down [01:53] on the nic, if I susspect a bad nic, I just replace it and see if it fixes the issue [01:54] it's a Dell PowerEdge server nic. [01:54] wth. [01:55] most dell poweredge servers have 2 ports on the back, I would try to plug in your nic to the second one, [01:55] make sure to set the same ip to it [01:57] If I remember back a while correctly, I think I ran into an issue once on one of my ubuntu desktop's that I set a static ip in after a dynamic ip that the route would keep getting erased on me. [01:58] lucasreborn_: that "task blocked for 120 seconds" stuff is usually caused by disk problems [01:58] lucasreborn_: though, are you using NFS? [02:01] i'm using NFS to store website data [02:01] it's a problem? [02:02] a disk problem could cause the server freezes up completely, but needing just a networking restart to get it all working back again? [02:03] this server was runnning about 15 days without problem, and problems appeared when both disks was full, but i did a clean install now, and the problem persists [02:07] lucasreborn_: right, so NFS issues that are solved by a networking restart would also cause that 120 seconds error [02:07] but it's not a nfs issue. [02:08] the cp isn't touching any NFS mounts? [02:11] no, i don't run a cp task [02:11] the nfs mount is dedicated to tomcat, reading and writing. [02:12] lucasreborn_: ok.. either way.. I'd suspect NFS problems due to network problems [02:12] ok [02:12] lucasreborn_: are you hooked up to a switch with auto-sense for the port speed? [02:12] i'll umount the nfs filesystem, and work a bit without him [02:13] yeah, a crappy dlink des 1008d [02:13] i'm just the trainee.. [02:14] lucasreborn_: sometimes autosense fails with nicer switches.. it can help to specify the exact speed you want [02:14] I've always found that odd, but cheap switches seem to do a better job of autosensing [02:15] is NFS mounted -ohard or -ointr? [02:16] i don't have the exact mount cfg [02:17] i will try to verify from here [02:17] i'm at home [02:17] i think is intr [02:17] how to specify port speed? [02:17] it's a gigabit nic connected to a fast nic [02:17] maybe that difference is causing the freeze [02:19] lucasreborn_: pastebin the contents of /proc/mounts and /etc/nsswitch.conf on the client. [02:25] guys [02:25] thanks for helpinh [02:25] here are the logs [02:26] http://paste.ubuntu.com/544256/ [02:30] are you there guys? [02:31] hello [02:32] I have a little question [02:32] lucasreborn_: sorry we're all kind of in and out between other tasks. [02:32] who could help me [02:33] william_a: just ask, sometimes it takes people a while to respond [02:33] i'll be here waiting, ok? just don't forget [02:35] lucasreborn_: just looking at your paste.. it does look like the link is dying [02:36] I'm running my ubuntu server 9.10 without monitor. Sometimes when I want to look what happen, I need to plug the monitor. But after the monitor plugged, it shows blank screen. It works if I restart the server with the monitor attached. Maybe some configuration I missed? [02:38] lucasreborn_: did you try just putting a new cat5e cable in? [02:38] lucasreborn_: I'd try ethtool -s eth0 speed 100 duplex full [02:39] if it stops dropping, then you'll want to run that on ifup [02:41] ooo mysql 5.5 GA!! [02:42] lucasreborn_: I'm off ... good luck [02:47] There must be a better way than that [02:47] ip link set maybe [02:49] hmm [02:49] any other suggestions? [02:52] william [02:52] did you just try to press enter when you attach the monitor? [02:52] yes [02:52] enter [02:52] numlock [02:52] everything [02:53] hmm [02:53] i don't know anything about server [02:53] i would boot my system with monitor attached and when needed reattach [02:53] this way the monitor is "pre-loaded [02:54] william_a: that's just how PC video works, you can't do anything about that. [02:55] it's a true server? [02:55] william_a: the BIOS is saying something like "hm, no monitor is plugged in. Therefore I will disable this video output." [02:55] i remember a option in bios postback video after resume, or something like this [02:55] It depends on the BIOS, of course. [02:56] indeed. [02:56] sorry guys for bad english [02:57] self taught [02:57] I remembered I can attach my monitor anytime with my older version of ubuntu [02:57] maybe it's different hardware [02:57] or a different driver/config [02:57] try to find in google [02:58] something that persists the driver loading etc. [03:01] can't stop mysql via 'sudo service mysql stop' and /var/log/mysql.err and mysql.log are empty. any idea about where i could start investigating the problem? [03:02] tried mysqld? [03:02] about twb reply, I'm thinking about dummy monitor that might cheat bios pretend that there is monitor plugged [03:02] or /etc/init.d/mysqld restart? [03:02] is there any? [03:02] sure there is (noob talking) [03:02] mysqld restart cant stop it too [03:03] william_a: that's called a KVM [03:03] Igoru: in ubuntu, log are stored in syslog [03:03] it tries to stop, fails. then tries to start, and says OK. [03:03] You can get a cheap four-port one for aroudn $20 [03:04] kvm.. it's software, right? [03:05] no [03:05] it's a hardware [03:05] you can manage multiple pc's with one mouse, keyboard and monitor. [03:05] it's a "switcher" [03:06] william_a i tried $ dmesg, but there's nothing about mysql there [03:06] igoru: try cat /etc/log/syslog | tail [03:07] william_a this file doesnt exist o_O [03:07] i'm running ubuntu desktop 9.10 [03:08] it's /var/log/syslog [03:08] I'm sorry, yes it's /var not /etc [03:08] thanks lucas [03:09] Dec 16 01:08:55 igoru-ubuntu kernel: [87939.007575] type=1503 audit(1292468935.492:68): operation="open" pid=25051 parent=25050 profile="/usr/sbin/mysqld" requested_mask="r::" denied_mask="r::" fsuid=0 ouid=0 name="/sys/devices/system/cpu/" [03:14] please [03:14] pastebin in http://paste.ubuntu.com the cat /var/log/syslog and /var/log/dmesg last lines. [03:15] william_a, the problem is not the monitor, it's the screen saver [03:16] if you are using a PS/2 keyboard, you have to have one plugged in while the system boots or it won't work [03:16] sorry lucasreborn_, http://paste.ubuntu.com/544268/ [03:16] the workaround is leaving it in all the time, or using a USB [03:17] qman_: the keyboard is plugged all the time [03:18] what sort of graphics hardware are you using? onboard/discrete, VGA/DVI/etc [03:18] it could be initializing the wrong connector or device [03:18] In the openvpn.conf file it says that the listen address is optional if i do not specify an address as the ubuntu server docs suggest what is the effect will it listen on all interfaces specifying an address becomes a problem when using dhcp [03:19] it's a PCI Express card VGA [03:19] set your dhcp to lease always the sampe IP [03:19] not sure about the brand [03:19] william_a, does your system also have onboard graphics? if so, make sure it's disabled [03:20] Igoru, these are all the lines? [03:20] also, does your keyboard's indicators light up with the numlock, capslock, scroll lock keys? [03:21] lucasreborn_ yeah... before those there are just repetitions, 'cos i tried the commands a couple of times [03:21] those line keep repeating for every time i try to stop mysql [03:21] jasonmchristos, that could be worked around with the firewall, just listen on all interfaces but configure the firewall to only accept connections from the desired interface [03:21] New bug: #690924 in mysql-dfsg-5.1 (main) "package mysql-server-5.1 (not installed) failed to install/upgrade: subprocess new pre-installation script returned error exit status 1" [Undecided,New] https://launchpad.net/bugs/690924 [03:21] qman: when I've plug the monitor and it shows blank screen, I always tried to press numlock key to see if the server crash, the light is responded well [03:22] did you mess up the logins etc? [03:22] sounds like this [03:22] i'm a begineer [03:22] i'm afraid i can't help you, sorry. [03:22] haha ty anyway, lucasreborn_ [03:22] Igoru, sound like a permission or something [03:22] i'm a kind of noob too. and i cant imagine how i could mess up with login if i'm trying to stop the service. === pgraner-afk is now known as pgraner [03:23] and i'm always using sudo. [03:23] did you installed mysql via apt-get or build manually/ [03:23] apt [03:24] i had this kind of problem some time before, when updating mysql [03:24] the solution was killing mysql process, updating, and then restarting it [03:24] but the database works? [03:24] william_a, then it's definitely the graphics, somehow [03:24] yes yes [03:24] i use it everyday [03:24] not sure if it's using the wrong one or if it's trying to be smart and not initialize because there is no monitor [03:24] would be fixed with a KVM switch [03:25] he could force initialisation ? [03:26] New bug: #690925 in mysql-5.1 (main) "Package MySQL 5.5.x for Natty" [Undecided,New] https://launchpad.net/bugs/690925 [03:26] but when you kill mysql process probably you're disconnecting something is connected. [03:26] I haven't experienced this problem, and I do basically the same thing on several machines [03:26] igoru: not sure, but you could check here: https://bugs.launchpad.net/ubuntu/+source/mysql-dfsg-5.1/+bug/448656 [03:26] lucasreborn_ what do u mean? i killed it with htop and then started with service start [03:26] Launchpad bug 448656 in mysql-dfsg-5.1 "CPU information is inaccessible for MySQL (dup-of: 444479)" [Medium,Triaged] [03:26] Launchpad bug 444479 in mysql-dfsg-5.1 "missing apparmor access rule " [Medium,Fix released] [03:28] igoru, check that link [03:28] it's the problem you're having [03:28] change the permission and owners of mysql using folders by chown [03:29] lucasreborn_ william_a reading, i think it will solve my problem. [noob speaking] but what is apparmor? [03:29] same question here [03:29] i'll search and post asap [03:29] lucasreborn_, your issue looks like a hardware problem of some kind to me [03:29] bad switch, cable, NIC, NIC driver, somethign [03:29] https://wiki.ubuntu.com/AppArmor [03:30] if you notice, eth1 keeps coming online right before/after your NFS drops [03:30] this problem appeared after 15 days of uptime [03:31] i'm really thinking it's that damn switch [03:31] I'm out... thanks for the help everybody and good bye.. [03:31] this shit appeared suddenly [03:31] good night william [03:32] g'night and thanks william_a! :D [03:32] qman__: ok very good i didnt want to specify the ip anyway as i am on dhcp [03:32] lucasreborn_, don't ignore the possibility of the NIC though, my file server's onboard NIC suddenly failed recently [03:33] of course it was more obvious, since the forcedeth driver was filling the logs with garbage [03:33] hummm [03:33] but they don't always go that way [03:33] two noob questions [03:33] how to change the eth number? [03:33] test one thing at a time, figure out what it is [03:34] example, i want my eth1 became eth8 [03:34] /etc/udev/rules.d/70-persistent-net.rules [03:34] there i can remove a eth too? [03:35] that is where the auto generated MAC to eth# associations are stored [03:35] example: i have a eth0 onboard. [03:35] added a nic, eth1 [03:35] it's pretty self explanatory when you look at it [03:35] changed the nic, eth2 [03:35] i want to delete the eth1 and make eth2 become eth1 [03:35] yes, that's where it is done [03:35] perfect! [03:36] i appreciate a lot your help [03:36] here it's 01:36 and i need to go to bed. [03:36] a lot of work to do [03:36] good night y'all [04:30] the openVPN server docs says to add this to the openvpn.conf but doesnt explain its use push "route 172.18.100.1 255.255.255.0" [04:31] will someone please explain this too me? [04:34] is this the gateway in regular networking? [04:35] of "default route"? [04:41] jasonmchristos, no, that is giving the VPN clients routes to your network [04:42] the clients already know where the internet is, but they don't know your network, only the VPN server [04:42] so you push routes to tell them where your network is [05:06] New bug: #690937 in bind9 (main) "package bind9 1:9.6.1.dfsg.P1-3ubuntu0.2 failed to install/upgrade: subprocess new pre-removal script returned error exit status 1" [Undecided,New] https://launchpad.net/bugs/690937 [08:06] What sets $FROM_SINGLE_USER_MODE in /etc/init/rc-sysvinit.conf ? === alvin_ is now known as alvin [08:21] New bug: #690983 in mysql-dfsg-5.1 (main) "package mysql-server-5.1 5.1.41-3ubuntu12.8 failed to install/upgrade: il sottoprocesso vecchio script di post-installation ha restituito lo stato di errore 1" [Undecided,New] https://launchpad.net/bugs/690983 [08:22] twb: Wait.. I know this one.... [08:22] Never mind, it was rcS [08:23] I wanted to throw away rc-sysvinit.conf because it reads /proc/cmdline, which is wrong for my containers (/proc/cmdline isn't virtualized), but it turned out to be too much trouble [08:23] It means if I start my VM server in single mode, then leave it and continue booting, all the containers will also try to enter single mode [08:28] So the question is.. do we a) have special boot code for containers, or b) virtualize /proc/cmdline ? [08:28] seems like there could be compelling arguments either way [08:35] The VM people say "you should not treat us specially" [08:35] The upstart people say "if you were doing your job, we wouldn't have to" [08:36] viz. detecting when you're inside a container and behaving differently, a la $debian_chroot handling in /etc/profile [08:36] well the ubuntu people sit in the middle and choose which one to snub [08:36] I think Ubuntu is pretty much in upstart's bed [08:36] sucking on its toes [08:36] Upstart is just so exotic.. we can't resist. [08:38] Is discussion about replacing upstart even possible? [08:39] that discussion is most likely going to come up a few times between now and 11.10 [08:39] I'm happy to hear that. [08:40] Even if we don't choose systemd or another boot "thingy" ... upstart's development has started picking up pace again. [08:40] alvin: what don't you like about upstart? [08:41] SpamapS: I don't know the ideas behind Upstart very well, but in it's current state, it is not production ready. It's the cause of a lot of boot problems. I can give examples. [08:42] alvin: that would be immensely helpful [08:43] I think most of the problems arise from poorly selected start on / stop on criteria [08:43] NFS mounting (in /etc/fstab) doesn't always work since upstart. Quasselcore starts before it sees a database, mdadm can't assemble drives and libvirt starts before its netfs drives are found. Sometimes nmbd doesn't start, and sometimes nfs-kernel-server starts before portmap,.. All in all, just 'booting' or rebooting a server has become imposible. You have to reboot some daemons after the normal boot in order to have the correct boot [08:43] order. That is very irritating. [08:44] (All those bugs are in Launchpad, and I experience them all) [08:45] alvin: I've seen a couple of them [08:46] At least 'file serving' should be a primary goal of Ubuntu-server. Something that should work at all times. It's not a very exotic feature. [08:46] and a few are also caused by missing post-start scripts to ensure that a service is *ready* before its 'started' event is emitted [08:47] agreed, but it's still missing and interrupting production. How can one take Ubuntu on the server serious then? The other problem is: someone needs to create these scripts and only needs to create them for Ubuntu? I don't know if the quasselcore devs will care. After all, only Ubuntu has troubles with starting their app and the database in the correct order. (for example) [08:47] One could argue that this actually MIGHT have affected sysvinit as well... as most sysvinit scripts return as soon as the daemon has forked.. but because everything is in serial.. the race between fork() and listen() and connect() rarely caused problems. [08:48] alvin: Are these bugs just not getting attention? [08:48] Yes, sysvinit was only flawed in theory. So, it's a good things replacements are made. But now we have taken some steps backwards. [08:49] I'm kind of surprised that upstart causes mdadm issues.. thats pretty early in the boot. [08:49] SpamapS: I don't think they do. Let me search the numbers a bit. [08:49] SpamapS, i don't see how it could if it is handled at the right time [08:49] The mdadm thing might be mountall or so. I don't know the details. I do know that I have to auto-assemble my RAID everytime after boot. Soemthing that should happen automatically during boot. (after all, it works automatically afterwards) [08:50] my rootfs is a mdadm raid1, never had an issue [08:50] alvin: mdadm has been really, really neglected in ubuntu server for a while. [08:50] alvin: the kernel should be detecting and assembling your raid partitions... [08:50] c0nv1ct: It's not rootfs in my case. [08:51] mdadm doesn't get involved unless there are problems. [08:51] alvin, it shouldn't matter really, they can all be assembled at the same time [08:51] the kernel should really be handling that, not your init [08:52] mdadm stuff: bug 27037 and bug 599135 [08:52] Launchpad bug 27037 in linux "mdadm cannot assemble array as cannot open drive with O_EXCL" [Medium,Fix released] https://launchpad.net/bugs/27037 [08:52] Launchpad bug 599135 in mdadm "mdadm cannot assemble array" [Undecided,New] https://launchpad.net/bugs/599135 [08:52] Quasselcore: bug 612729 [08:52] Launchpad bug 612729 in quassel "quasselcore does not connect to database at boot" [Undecided,Expired] https://launchpad.net/bugs/612729 [08:54] NFS: bug 540637 [08:54] Launchpad bug 540637 in nfs-utils "nfs-kernel-server fails to start after kernel upgrade" [High,Triaged] https://launchpad.net/bugs/540637 [08:54] and bug 540637 [08:55] hmm, havent ran into that one, then again i do have a pending reboot for a kernel update [08:55] no, scrap that last one [08:55] alvin: with mdadm.. I really don't think upstart is at fault for your drive disappearing. [08:55] It doesn't always happen. The solution is: service nfs-kernel-server stop; service portmap stop; service portmap start (yes, no restart) and service nfs-kernel-server start [08:56] alvin: unless you're using some kind of hotplug [08:56] SpamapS: No hotplug. Upstart might not be at fault. Still, this has worked and no longer does. [08:57] I even took a backup and recreated the RAID [08:57] alvin: Right, the mdadm user space tools were allowed to lag WAY behind upstream for some reason [08:59] alvin: looking at the NFS upstart jobs [09:01] I'm happy that someone is looking into it. I had the feeling that NFS was being neglected (not only in Ubuntu) in favour of Samba for file sharing between linux machines. [09:02] alvin: IMO, they accomplish entirely different things [09:02] n [09:02] Agreed [09:04] alvin: right, so the quassel problem, btw, is caused by /etc/init/rc-sysinit.conf .. This is one that drives me nuts too.. net-device-up IFACE=lo means that you may not have eth0 up yet. [09:04] and that's a problem, because the database is on another (virtual) server [09:05] So, quassel doesn't see the database and asks the first user to create one. [09:06] alvin: right, I haven't come up with a good generic solution to that. The workaround is to change rc-sysinit.conf to bring up init.d scripts on IFACE!=lo [09:07] I see how that could work. Now I'm thinking where there could be trouble with that scenario [09:09] alvin: because its an AND with filesystem.. there is one danger.. that the ifup script will get stuck waiting forever [09:09] So, if your network card can't go up. You're screwed [09:09] alvin: but no, ifup doesn't wait for net-device-up [09:10] alvin: so yeah, it should be fine [09:10] alvin: actually I think the NFS bug might be the same issue [09:11] Yes, and it's definately worth a try. Thanks. I can't test now because too many users are using the servers, but I'll write it down for the next reboot. [09:11] actually [09:11] portmap will always come up before runlevel2 is triggered [09:12] virtual-filesystems always happens before filesystems [09:13] Ah, but I would want libvirt to wait for netfs filesystems before trying to start (and fail) machines that use those netfs filesystems. I forgot to give the bug number. Let's see [09:13] virtual-filesystems is /proc, /sys, stuff like that [09:13] bug 491273 [09:13] Launchpad bug 491273 in libvirt "netfs storage pools are not autostarted at boot (karmic) (dup-of: 351307)" [Undecided,New] https://launchpad.net/bugs/491273 [09:13] Launchpad bug 351307 in libvirt "Libvirt NFS mount on boot." [Low,Incomplete] https://launchpad.net/bugs/351307 [09:13] SpamapS: oops, ok [09:13] it goes virtual -> local -> remote -> all [09:14] alvin: btw, THANK YOU for reporting bugs like this. :) [09:14] Remote fs can also be virtual :-) in another sense. [09:15] yeah, I'm just a user (of a lot of different servers). Sometimes I wonder why other people do not run into stuff like that. (Well, they do, but not as much as I would think) [09:16] alvin: well , know that its appreciated.. and we're trying as hard as we can to get stuff fixed in new releases and push what fixes we can into the stable release. [09:18] I'm glad it's appreciated and that someone takes care of the reports. Should I change the 'expired' status of the Quasselcore one to 'new'? [09:19] the libvirt one doesn't make much sense either actually [09:19] libvirt starts at runlevel 2.. which is triggered by filesystem + net-device-up IFACE=lo ... [09:19] oh wait that makes perfect sense. ;) [09:19] Yes :-) [09:20] hmm.. tho I thought filesystem was only triggered when all (remote, virtual, and local) were mounted [09:20] * SpamapS is reading mountall's code and doesn't understand [09:21] libvirt takes care of mounting itself. It's called 'storage pools' [09:22] alvin: ah, so the NFS filesystems might not be auto-mount ? [09:22] You can also mount NFS drives in /etc/fstab and tell libvirt it's a 'dir storage pool'. The problem is that if your network is a bit slow in coming up (don't know how to tell it), your NFS drives might not be mounted. Now, this problem is better, but I still see it from time to time. [09:22] SpamapS: True. libvirt actually does the mounting when the daemon starts. [09:22] alvin: so then its possible that filesystem triggered before eth0 came up [09:23] Exactly [09:23] I haven't tried iscsi. Libvirt can manage iscsi as well [09:23] alvin: this is the focus of something I'm proposing right now actually.. [09:23] alvin: isn't that an upstart issue? [09:23] alvin: that we have one file.. /etc/init/network-services.conf .. that defines when everything else gets started.. [09:23] Actually, that happens every time. The network is never up before libvirt. [09:24] i.e. just tell the libvirt job to depend on the network-filesystem job [09:24] alvin: but, the workaround would be to have libvirt start on net-device-up IFACE=eth0 [09:24] twb: maybe it is, yes. But it's only an issue if you actually have remote storage [09:25] SpamapS: yes (if you have remote storage) [09:25] I wasn't suggesting you'd make that change for everyone [09:26] No, but can upstart know the difference? [09:26] alvin: right, so change it from runlevel [2345] to net-device-up IFACE=eth0 ... or started networking if its a static interface. [09:28] I'll have to set the static interfaces back one day... All my virtual servers use dhcp because of bug 446031, but that is no longer necessary. [09:28]