[01:35] kirkland: still there? === swift_ is now known as swift === swift_ is now known as swift [01:49] heh! [01:49] i installed nginx with aptitude install nginx... and removed it with aptitude --purge remove nginx... but guess what? /etc/init.d/nginx is still there, as is /etc/nginx and all the manpages etc [01:58] steven_t: file a bug then [01:59] lol [02:10] ScottK: sup? [02:11] lamont: I was thinking about the new postscreen tool in 2.7. [02:11] Upstream has clearly labled it "Experimental" in 2.7. [02:11] Should it be split out into a separate binary that's not installed by default? [02:12] hrm... possibly [02:12] I was thinking that experiments shouldn't be part of the default mail server task for an LTS. [02:12] I would not be averse to such a thing [02:12] ah, come on.... where's your sense of ADVENTURE?? [02:12] er, I mean, I agree [02:12] Heh. [02:13] you wanna work up a diff? [02:13] I know that 'experiment' in the default install is now an Ubuntu Desktop tradition, but for Server, I think not so much a great idea. [02:13] harsh dude [02:13] Accurate. [02:15] OK. Let me see what I can do. [02:17] hmmm [02:17] Oh, how is spam filtering in the mail task going? [02:19] I lost track of how far we got on that. [02:19] ivoks would know, but he's not around. [02:20] In any case, amavisd-new with spamassassin and clamav is the standard, documentation approach. [02:22] awesome [02:22] I've been trying with dovecot-antispam, because it seems it would give the best result, if I could get it working (: [02:25] (I.e monitors a spamfolder, and on movement out of the folder, calls the spamfilter automatically to mark as 'ham') [02:26] However, documentation on it is nonexistent :( [02:27] cloakable: Start with the Ubuntu Server Guide documentation on spam filtering. [02:28] dovecot-antispam would be an advanced part you might bolt onto it later. [02:28] ScottK: That's a little clunky to train for nonspam, though. [02:28] Needs ssh-ing into the server to call manually. [02:29] And while I can do that, I'd rather not have to, and there's users on my server that cannot :) [02:31] With a good set of RBLs + amavisd-new/spamassassin/clamav you get rid of an awful lot of it without having to mess with bayesian filter training. [02:31] I'd get that set up first and then see if you want to bother. [02:34] Mmmmm. [02:34] And when I get spam in my ham and ham in my spam? :P [02:35] First see how much of it is before you solve the problem. [02:47] * cloakable finds out what was wrong with dovecot-antispam >.> [02:48] It's been compiled with the wrong backend :) [02:49] * cloakable gets the source, comments out 'mailtrain' and puts in 'dspam' === cyphermox_ is now known as cyphermox [03:22] There, added 'dovecot' to the dspam trusted list [03:32] New bug: #559745 in eucalyptus (main) "NC failed to start a session with a libvirt internal error" [Undecided,New] https://launchpad.net/bugs/559745 [03:41] New bug: #559752 in samba (main) "package samba-common 2:3.4.0-3ubuntu5.6 failed to install/upgrade: el subproceso script post-installation instalado devolvió el código de salida de error 1" [Undecided,New] https://launchpad.net/bugs/559752 [03:49] lamont: Never mind. Apparently it's so experimental Wietse didn't include it in the tarball. No wonder I couldn't find it. [04:04] New bug: #553853 in samba (main) "(Kubuntu) Samba shares (fstab) slow down system shutdown/reboot" [Undecided,New] https://launchpad.net/bugs/553853 [04:11] ScottK: he tends to be more pedantic than me about experimental vs official - which lets me ignore that aspect without thinking much about it [04:11] to the point that you made me go "wut, yeah kill that" thinking you'd actually seen it there already [04:11] Which is a good thing to have in an MTA author. [04:12] very much so [04:12] on that note, sleep time. [04:12] I thought I'd seen it referred to as being in the release onthe mail list. [04:12] head->pillow [04:12] Good night. [05:50] I just killed two birds [05:50] good hello [05:51] I wish I would use the right channel [06:01] It's more fun for us when you don't. === swift_ is now known as swift [07:00] hey, i added a scsi drive to my live server and it isn't showing up, do i have to reboot it? === swift_ is now known as swift [07:10] ZenMasta_: are you there? [07:11] need some help installing pdo and pdo_mysql i get a message sh: phpsize not found [07:12] looks like pdo is part of the php code now and you should just need to skip to pdo_mysql [07:12] i see, let me try and see what happends [07:13] histo same error [07:13] ZenMasta_: yeah what version of ubuntu are you using? [07:13] 9.10 [07:14] on a side note, when I try to install pdo_mysql it downloads pdo_mysql and then after it downloads pdo still [07:14] and you're using sudo pecl install pdo [07:14] yep [07:15] do you have php5-dev installed? [07:16] I think so how can i find out without trying to install it again [07:16] dpkg -l | grep php5 [07:17] just decided to install before you typed that [07:17] should show a php5-dev package but like I said i think pdo is obsolete [07:17] so we'll see what happends [07:18] yeah pdo has been moved into the php source [07:18] histo that did it [07:18] installing now so i'll try the web app when its done and hopefully it will progress [07:18] did you try the webapp prior to running pecl [07:19] !info php [07:19] Package php does not exist in karmic [07:19] !info php5 [07:19] php5 (source: php5): server-side, HTML-embedded scripting language (metapackage). In component main, is optional. Version 5.2.10.dfsg.1-2ubuntu6.4 (karmic), package size 1 kB, installed size 20 kB [07:20] ZenMasta_: http://pecl.php.net/package/PDO/php-src/pdo [07:20] see [07:21] thanks === swift__ is now known as swift === swift__ is now known as swift [07:51] how do I edit php.ini? when I try to open it with vi it's as if it doesn't exist so it pretends to make a new file [11:59] Hi [13:10] Hello - i would like to test out the ubuntu enterprise cloud with more than 1 or 2 nodes, so i was considering if you can run UEC on Amazon Ec2 for testing? It seems this is the only way to "rent" a lot of computers for a short period of time [13:32] hi all. seems I'm doing something strange here. I try to mount an nfs filesystem on the host from a virtualbox VM, but I get 'mount.nfs: mount to NFS server 'rpcbind' failed: RPC Error: Program not registered' - 'mount.nfs: internal error' - /etc/exports looks right, services are started and ufw has 'allow from x.x.x.x' (the VMs address). The VM runs in bridge mode. Any ideas? [14:05] RoyK^: is mountd running on the server? [14:06] hm. no. what starts that? thought that should be in nfs-kernel-server or something [14:08] Start by checking "rpcinfo -p" [14:14] http://pastebin.com/98hsiw0x [14:14] KristianDK: You could use public cloud from Eucalyptus [14:15] binBASH, but i want to check out the configuration :-) Not how the instances works [14:15] Isn't the whole point of EUC that it's backwards-compatible with Amazon? [14:15] Er, UEC [14:16] KristianDK: It's possible to run everything on one node. [14:16] Not really a need for multiple nodes....... [14:16] New bug: #560011 in ntp (main) "Time cannot be fixed with ntpdate" [Undecided,New] https://launchpad.net/bugs/560011 [14:16] binBASH, both controller and node in one box? [14:16] yeah [14:16] twb, pmatulis any ideas? [14:16] have the same here [14:17] RoyK^: pastebin "exportfs -vra" [14:17] KristianDK: I have 7 nondes, including the one with cluster and cloud controller [14:18] twb: exporting 213.236.233.67:/var/www [14:18] planning to have 150 nodes ;) [14:18] tried with * as well [14:18] binBASH, cool :-) Well, i want to test things out before deploying it in a big scale [14:19] Like me then ;) [14:19] binBASH, do all your nodes have the VT extension as its recommended? [14:19] yup [14:19] KristianDK: http://www.hetzner.de/en/hosting/produkte_rootserver/eq6/ [14:20] have those as nodes [14:20] 150 nodes??? how many racks? [14:20] binBASH, i was actually considering http://www.hetzner.de/en/hosting/produkte_rootserver/eq8/ :-D [14:20] im already a customer there for some other servers [14:21] KristianDK: It will be a problem with network configuration ;) [14:21] still stucked at this...... [14:21] yeah, i guess because of the IP adresses being bound the MAC and the limitation of the 100mbit router, right? [14:22] Yeah, the ips are bound to server. [14:22] RoyK^: Dunno, this providers does tower hosting...... [14:23] KristianDK: With 150 nodes, there will be a lot of ram anyways, don't need eq8 really ;) [14:23] binBASH: if you need 150 nodes, I'd guess hosting it locally may be a lot cheaper [14:24] RoyK^: Don't think so. [14:24] don't have money to buy all those servers ;) [14:24] binBASH, true :) I think i'll end up with an ESXi more on the EQ8 anyway, since everything else seems complicated with hetzner :( [14:24] binBASH: but .... 150 nodes? you can run like 10 VMs on a single node - perhaps more - what do you need this for? [14:25] KristianDK: Well, I'll try to start vms now with vnc option and configure networking inside there manually. [14:25] RoyK^: Need them not for the vms, but for storage [14:25] binBASH: how much storage do you need? [14:26] binBASH, i've configured a router VM which the IPs are bound to, it forwards the IPs [14:26] RoyK^: 200 TB [14:26] binBASH: I just got an offer for such a box - NOK 250k [14:26] KristianDK: I don't want to NAT, because 2 TB Traffic Limit per node [14:26] binBASH: and storage should be done on zfs imho [14:26] NOT on a VM [14:26] but on hardware [14:27] RoyK^: I'll use GlusterFS [14:27] why not just a big supermicro box stuffed with 2TB drives and a SAS expander and some extra chassises for disks? [14:27] RoyK^: Like I said, I'm limited in Finances ;) [14:27] it'll be cheaper [14:28] binBASH, i dont use NAT, its Ip forwarding, i use the router as gateway in the network config - but i don't think you can get around the 2tb limit anyway? Its bound to the IPs [14:28] you need 200TB and can't afford it? [14:28] RoyK^: It's a single point of failure. [14:28] binBASH: then get two of them and use zfs send/receive to keep them in sync [14:28] KristianDK: if you forward to one server it will count there [14:29] RoyK^: Atm I'm using 20 TB Raid 6 NFS Server. [14:29] binBASH: I REALLY doubt you can get something cheaper from somewhere else [14:29] binBASH: we bought this box some time back with 30TB net storage - it just cost like USD 10k [14:29] RoyK^: Everything I was looking for was more expensive. like NetApp or Isilon [14:29] binBASH, as i understood from hetzner you need one router VM per physical server [14:30] binBASH: hah - use supermicro hardware, cheap drives, and opensolaris with zfs (with compression and dedup) [14:31] binBASH: where I work, we have rather high storage demands - windfield and other satellite data takes up space, and we're getting more and more all the time [14:31] KristianDK: If it would be like this, I wouldn't use the vm for routing. [14:32] Would just use the main box itself, because the ip is unusable anyways from within vms. [14:32] binBASH, i was told this was the only option - what else would you do? [14:32] true [14:32] i was thinking ESXi again [14:32] sorry :P [14:32] KristianDK: I already, started vms manually and they had a usable ip address. [14:32] but I dunno how to automate it within Eucalyptus, that's the only problem [14:32] yeah [14:33] binBASH: really, using VMs for storage is a BAD idea [14:33] RoyK^: I don't wanna use vms for storage ;) [14:34] the storage will be on the real server itself, though I will embedd it from inside the vms [14:34] binBASH: but - please - give opensolaris+zfs a try - it's well worth it. no raid controller, just zfs doing it all [14:34] zfs rocks rather loudly [14:35] RoyK^: If I would use opensolaris on some boxes I wouldn't be able to use their processors. [14:35] I have rather high demand on cpu [14:35] storage speed is not that important [14:36] for your needs, I would say separate storage and computation [14:36] use storage computer cpu for compression and dedup [14:37] Don't need compression [14:37] depending on the data, both can give you quite a bit of gain without much cpu use [14:37] for jpgs it's useless [14:37] what sort of data is this? [14:37] indeed [14:37] RoyK^: We're hosting image agencies. [14:37] Things like www.gettyimageslatam.com [14:37] but stuff like zfs snapshotting is quite priceless [14:38] btrfs and LVM do snapshotting [14:38] btrfs is NOT stable [14:38] Granted [14:38] LVM snapshotting is crap [14:38] LVM snapshotting is adequate for my purposes [14:38] LVM snapshotting moves data out of the original place for each write instead of writing new data and moving pointers [14:39] meaning if you have lots of snapshots, everything will be very, very slow [14:39] Um, both LVM and ZFS snapshotting are block COW. [14:39] I grant you that LVM is probably a lot slower. [14:39] lvm moves data out before overwriting them - not like zfs, which writes new data [14:40] Shrug. [14:40] CoW is two different things - either write new data and move pointers, which is what ZFS and NetApp does, or move the old data prior to overwriting the old ones, which is what LVM does [14:40] maybe I'll take Strato HiDrive Pro for Storage ;P [14:40] At the end of the day, ZFS is not enough to make me adopt osol. [14:40] 5 TB mirrored = 149 Eur / Month [14:41] twb: heh - then you really haven't looked into it [14:41] I'm running a 2TB osol server for ZFS right now. [14:42] RoyK^: If you really can afford a big storage you should go to Isilon ;) [14:42] But as soon as btrfs is ready, it will die [14:42] It's a much better technology [14:43] btrfs is decent, but lacks a lot of what's in zfs atm. give it a year or two and it might catch up [14:43] but yes, I will also switch to btrfs once it's there [14:43] Exactly [14:44] Which will unfortunately not be until 2012 (for LTS) :-( [14:45] KristianDK: I really don't know how to master that network problem ;) [14:45] but then, I can't wait two years for a storage solution, and then opensolaris is the way togo [14:45] binBASH, i think we need to talk to hetzner [14:45] they are kind of blocking for allowing this [14:45] :p [14:46] KristianDK: Like they're blocking gigabit as well [14:46] exactly, i asked them for gbit [14:46] and you can actually get that [14:46] KristianDK: You can have it, just costs........ [14:46] yep [14:46] you need to have flexipack and another nic [14:47] and additionally a switch [14:47] and additionally 69 Eur for moving all your servers so they are beside each other. [14:49] binBASH: what makes me wonder is why you (or your company) are hosting terabytes of data and can't afford a decent (and quite cheap) storage solution, like the osol-based one we have. It can be expanded quite easily with a SAS expander and won't cost a lot using WD Green drives or so [14:50] RoyK^: Because agencies don't pay that much ;) [14:51] RoyK^: And they pay monthly. Not a year in advance :p [14:51] well, loan some money and it'll pay back quite quickly [14:51] for EUR 10k, you get 30-35TB, which I guess will be sufficient for some time [14:52] RoyK^: Like I said we have already 20 TB. [14:52] no, wait, more - wait... [14:52] lol [14:52] and we bought it for 9K 2 years ago [14:52] though it's a single point of failure and not mirrored [14:53] hi leonel [14:54] ea binBASH [14:54] binBASH: just got this offer - supermicro box with 36x2TB disk and an ok motherboard, a bunch of memory, some cpus etc, meaning if you use three RAIDz2 groups of 12 drives each, it gives you 20x3=60TB -> price NOK 86k, around EUR 10k, and possibly cheaper outside Norway [14:55] 34 2TB drives, that was, but still (forgot about the root SSDs) [14:55] 10K for one box I assume ;) [14:55] yes, but it's still cheap [14:55] Here we pay 8500 Eur for a box with 24 x 2 TB [14:55] with 3xraidz2, you can loose six drives in total [14:56] wierd - that's _more_ expensive :) [14:56] * RoyK^ thought Norway was meant to be the expensive place [14:56] though if a box fails raid is useless ;) [14:57] binBASH: then get two and use zfs send/receive to mirror the two [14:57] and when storage fills up, get a SAS expander and an extra chassis and some drives [14:58] osol not supporting ext2 is a real pain in the arse. [14:58] over 3-5 years, I would guess you would save LOTS of doing this yourself instead of paying others to do the same [14:58] twb: why should it??? [14:58] So that I can seed the osol box by sneakernet instead of our shitty 100baseT and ADSL lines [14:58] RoyK^: Well I would lack then cpu power for video processing [14:59] binBASH: don't! [14:59] binBASH: use NFS [14:59] or iSCSI [14:59] or CIFS [14:59] RoyK^: We're having NFS already ;) [15:00] nfs performs well enough for that - for those storage needs, it would be silly to put all services in one place [15:00] huh? [15:00] get a storage server with sufficient memory and cpu for the storage alone and get compute nodes to do the ugly stuff [15:01] RoyK^: what's his use case? Just normal office documents and such? [15:01] twb: images and video [15:01] I think getting 150 Nodes which offer 1200 cpu cores + 200 TB Storage is better ;) [15:02] if planning for 3-6 months, sure, but if you are planning to be in business for a long time, buying hardware will save you a lot of money [15:02] Yeah, NFS over 1000baseT to a single NAS or SAN is probably the Right Thing. [15:03] RoyK^: You forget the fact you normally throw out servers every 2 years [15:03] binBASH: not really - storage servers can last a LONG time [15:03] especially with zfs - autogrow is nice [15:04] take a zfs mirror, replace one part with a larger drive, resilver, replace the other, resilver, and zfs says 'oops - I'm bigger' [15:04] same with raidz volumes [15:04] Raid rebuild will take ages [15:05] not really - for 30TB a scrub takes a couple of days [15:05] I already takes 4 hours to rebuild the current raid ;p [15:05] resilver about the same [15:05] and replacing drives isn't what's done daily [15:05] but hey, I've just been working with storage for 10+ years, do as you please [15:06] I think a distributed storage architecture is much better. [15:06] Companies like NetApp or Isilon doing it as well. [15:06] how much do they charge you per month for 200TB? [15:06] binBASH, have you, btw, checked for hetzner alternatives? [15:06] KristianDK: Yup [15:07] KristianDK: But only in Germany [15:07] binBASH, and they have the same sucky setup? :P [15:07] binBASH, are you german? [15:07] They are even mor worse. [15:07] NetApp is doing quite well, yes, but they charge you EUR 100k for a few terabytes [15:07] KristianDK: I live in Switzerland :) [15:08] binBASH, ok - cool :-) I'm from Denmark, so i speak a bit German, but sometimes i really don't get what they are trying to tell me at hetzner :P [15:08] RoyK^: Things like GlusterFS are working like this. [15:08] KristianDK: I moved from Germany to Switzerland in 2007 [15:09] does that support stuff like versioning or snapshotting? [15:10] binBASH, well, the problem is i've been searching for alternatives to hetzner, but they seem remarkably cheap compared to everything else [15:10] and im actually satisfied with everything but their network setup :P [15:11] binBASH, however - they recently introduced the failover IP thing [15:12] which redirects and IP to another server [15:12] maybe we can work something out with this thing? [15:12] binBASH: but how much for 100T? [15:12] or 200 [15:17] New bug: #560047 in dovecot (main) "new upstream version available" [Undecided,New] https://launchpad.net/bugs/560047 [15:17] RoyK^: Like I said each node gives 2,7 TB [15:17] For mirroring you need 2 nodes. [15:18] A node costs 69 Eur / month [15:18] and provides 8 cpu cores which I can use for video rendering and image processing [15:19] because we're a swiss company we don't have to pay German VAT [15:19] so it's cheaper. [15:20] so it costs like 9500 Eur / Month. [15:23] KristianDK: For the failover ip you need flexipack, which is 15 Eur / Month [15:24] RoyK^: I would agree as pure storage it's too expensive. [15:27] good morning - I have a Dell Precision 650. I want to run SATA drives on it as boot devices. Can I install a 3rd party SATA PCI card and boot from a drive connected to it? thanks [15:28] You should be able to. [15:29] Absolute worst case scenario you unplug the installed drives from the built in controller, install, and then reconnect them. [15:29] ScottK: how do I get the BIOS to see the PCI card? [15:30] The one time I've had to worry about it, it just did. [15:30] (never booted from a add-in card) Cool - I'll give it a try - Thanks ! === martin- is now known as zx === zx is now known as martin- [15:33] RoyK^: http://gluster.com/community/documentation/index.php/Main_Page#Gluster_Filesystem [16:06] binBASH: :P [16:07] koolhead17: ? [16:07] gluster [16:07] koolhead17: It works here without problems so far. [16:07] binBASH: it rocks [16:11] koolhead17: are you using it? [16:12] binBASH: my friend owns the company behind this project :D [16:13] ohh :p [16:14] binBASH: he is the lead developer too :D [16:14] very cool [16:14] glusterfs is very good design I think. [16:15] Too bad I can use it with 100 Mbit only koolhead17 :p [16:15] binBASH: heh. poke them [16:15] i think #gluster [16:15] koolhead17: It's not a gluster issue, servers only have 100 mbit ;) [16:16] ScottK: absolute worst case is a wincontroller :-P [16:16] twb: True. [16:17] Or my boss's favourite trick -- buy a server with hotswap bays, but forget to buy the RAID5 chip for the hardware RAID controller [16:18] lol [16:18] In which case I could create up to two RAID0 arrays of one drive each, so I couldn't even make an md RAID5 [16:18] twb: Sounds more like epic fail than a trick ;) [16:18] binBASH: he has done it TWICE [16:19] twb: So he didn't learn? [16:19] And we're still running Pentium IIIs, so you can imagine how rarely we buy new gear [16:19] Pentium 3 omg [16:20] twb: How much people you are in company? [16:20] Probably about ten [16:21] ok, more than us then ;) [16:21] It's hard to tell because some spend months pimped out, and some ex-employees continue to lurk on the lists [16:21] We replaced the LaserJet 4 last month, and the sysadmin deploying it went "oh, cool, the NEW unit has only been EOLd by HP since 2008" [16:21] lol [16:22] Having said that, I loved that little LJ4 [16:22] sounds like a lack of money [16:23] There's a policy of handing most of the profits to the engineers instead of the company [16:23] But it's also a mindset thing. [16:24] We have a pair of Q9550 with 2TB of storage, but one got stolen to run rpppoe. [16:25] twb: We had such a policy as well twb [16:25] All money to personal [16:25] ;) [16:25] Get a project manager for 100K Eur / Year [16:25] kick him out 9 months later because he sucked [16:26] and have a second boss, which was also not very useful [16:26] * cloakable eyes gluster storage platform >.> [17:16] hi id like to set up monthly bandwidth quotas for my home network something like http://www.digirain.com/en/trafficquota-overview.html but ive been googling like crazy and i cant find anything like that for ubuntu, any suggestions? [17:28] pjp3rd: You can use iptables to set a quota [17:28] http://linuxgazette.net/108/odonovan.html [17:28] pjp3rd:http://linuxgazette.net/108/odonovan.html [17:29] pjp3rd: https://help.ubuntu.com/community/UFW [17:33] brianherman, thanks that looks like a good place to start [17:33] pjp3rd: why you want quote in home network? [17:34] pjp3rd: Use the ubuntu one it seems the simplest [17:34] Quota not quote [17:34] quota yeah ;) [17:35] brianherman, but id need to set up a seperate quota for each user, a way for user to check his quota and to automate it everymonth. so it would be nice if someone has already done that work.. [17:35] binBASH, cus the ISP gives me a quota and im sharing it with 16 people, seems like the most sensible way to avoid fights when we are running out after 2 weeks each month [17:36] pjp3rd: If you share the line with 16 people why not limit bandwidth? [17:37] pjp3rd: http://manpages.ubuntu.com/manpages/karmic/man8/tc.8.html [17:37] binBASH, im not sure what you mean? [17:37] pjp3rd: http://manpages.ubuntu.com/manpages/hardy/man8/tc-cbq-details.8.html [17:38] binBASH, im not looking to shape the bandwidth for speed im looking to make sure we dont exceed the monthly limit [17:38] ok [17:42] pjp3rd: there's something called "bandwidthd" that might help in some way. [17:43] pjp3rd:http://bandwidthd.sourceforge.net/ [17:44] sherr, bandwidthd helps half the problem if it can moniter the bandwidth. id like to enforce limit as well [17:44] pjp3rd: At least then all would know who steal the traffic :p [17:45] binBASH, yip it would tell me but too late rather than fighting with people every month id just like to divide it equally and no more worries [17:46] pjp3rd: if you limit bandwidth for everyone it will be equal [17:48] pjp3rd: Think limiting bandwidth is better than having no internet for half of the month ;) [17:48] k, based on brianherman's tip ive found www.linuxquestions.org/questions/linux-networking-3/iptables-to-stop-bandwidth-completely-592827 which is what im basically looking for [17:48] but your decission ;) [17:49] problem is it doesnt seem like such a polished solution [17:49] binBASH, im not sure what you mean [17:49] pjp3rd: You said you have a quota, what if you exceed it? [17:50] my isp enforces a quota when we exceed it the connection is throttled to basically unusable speeds [17:51] pjp3rd: So why don't you distribute a max. bandwidth equally? [17:51] binBASH, what do you mean? [17:51] I mean, I wouldn't accept the fact if I'm amongst the 16 people, and one causing so much traffic, so I would have no internet then for half a month [17:52] pjp3rd: With the tc links I posted, you can assign each user equal bandwidth. [17:53] so it's not possible traffic limit will be exceeded [17:55] binBASH, correct me if im wrong but what i can understand from the link you posted tc/cbq can shape my connection meaning how much is been used by any given user/protocol at a given time, thats not going to help me to stop total monthly usage from exceeding the isp quota is it? [17:56] pjp3rd: With that you can limit the bandwidth for each user. So every user has equal line. [17:56] and you can setup a max. bandwidth rule as well, so with that you can't exceed your providers traffic limit. [17:57] binBASH, oh i didnt see details about that? how can i set up a monthly limit? [17:58] pjp3rd: You don't set a traffic limit. You set a bandwidth limit. [17:58] The Linespeed will be slower though [17:59] binBASH, can you give me more details? [18:00] pjp3rd: http://www.oamk.fi/~jukkao/lartc.pdf [18:00] read this ;) [18:00] binBASH: do you really need like 5k cores for this? I thought you were doing storage [18:01] or 800 cores, that is [18:01] RoyK^: Storage and Processing ;) [18:02] 1200 cpus actually [18:02]