[02:17] is anybody here? [02:19] somewhat === RudyValencia- is now known as RudyValencia [04:41] can anybody tell me how I can make a user go into home/user/public_html when using sftp? Also how can I make sure that user can only see his directory? [04:42] TimR, set his home to /home/user/public_html [04:43] well see im using webmin to control everything [04:44] dunno webmin, but it should have a way to customize user homedir ! [04:47] TimR, http://library.linode.com/security/sftp-jails [04:48] well see only group setup I have is users [04:48] can i change that from filetransfers to users? [05:22] For some reason my server is loading extremely slow. Swapped out some ram and the CPU and even grub sits there for a while before it starts booting. [05:28] It does eventually boot though, and for some reason its still not mounting my array...... === iggi__ is now known as iggi_ [05:35] oh well still working for now at least lol [05:38] just annoying that I have to go manually mount that drive everytime I reboot [06:49] HI I have problems with my apache, I installed it using tasksel for lamp-server . It seems searching a non existent .htaccess [06:49] (13)Permission denied: /home/user/belajar/.htaccess pcfg_openfile: unable to check htaccess file, ensure it is readable === Major_Tom is now known as Omega42 [07:56] i need u r help to setup local repository in our office. We are recently migrated to Ubuntu 10.04. Currently every pc is updating from internet. (we are having 10 pc's in our office) [08:39] I want my Ubu Server 10.4.03 machine to act as a router/firewall between the internet on eth2 and the switch (and attatched machines) on eth1. What do I need to look up? === bennym_ is now known as bennym === bennym is now known as bennymAFK [10:16] Hey im trying to join ##Security on Xchat on windows but it does not add it to my chennel list does any one know why that might be? [10:45] Good morning Everyone! [10:47] is it possible to install JDK in EC2 while running from Natty image? [10:48] my instance hangs when I try to do so [10:48] both natty and Oneiric [10:48] please advice [11:06] hello! I have a server randomly rebooting on me, and while I am looking into what is going on, I noticed I have a 'kworker' process eating up quite a lot of resource. how can I find out what this process is doing, and why it is running so much on an otherwise idle server? [11:06] eg: 183 root 20 0 0 0 0 S 22 0.0 184:07.69 kworker/6:1 [11:06] 27003 root 20 0 0 0 0 D 3 0.0 1:15.80 kworker/u:2 [11:25] is drdb capable of rebooting a server when it should not, if it is misconfigured? [12:23] CrazyGir: maybe that's a feature of the cluster to stabelize the system when it's unclear who's the master and who's the slave. [12:24] (n.b.: I'm pulling this out of my ass from peripheral experience with Oracle and Sun clusters, I've never worked with drdb) [13:18] yea, I think you might be right [13:18] any thoughts on the first question RE kworker [13:55] Hi all [13:55] quick question: i'm wiping drives real quick, so i'm using dd if=/dev/random of=/dev/xxx. does this command end when it reaches the end of the drive, or does it continue forever? [14:01] hello [14:02] can anyone help me with a bind problem? [14:02] !ask [14:02] Please don't ask to ask a question, simply ask the question (all on ONE line and in the channel, so that others can read and follow it easily). If anyone knows the answer they will most likely reply. :-) See also !patience [14:02] ok thanks [14:04] when I do nslookup ubuntu.com the query reports server 10.61.3.245 address 10.61.3.245 is ther a way to make bind report server (my server name) [14:13] When I do nslookup ubuntu.com my bind server reports server 10.61.3.245 address 10.61.3.245 is ther a way to make bind report server (server name ) instead of the ip address [14:17] ljaclark, do you know what nslookup does? [14:17] it converts name to ip address [14:21] yes. I have asked the question because I have to setup bind for a uni project. The lecturer wants me to have the DNS server show the server name instaed of the ip address. in the first line of the nslookup command [14:22] root@group2:/etc/bind# nslookup www.howtoforge.org [14:22] Server: 10.61.3.245 [14:22] Address: 10.61.3.245#53 [14:22] Non-authoritative answer: [14:22] Name: www.howtoforge.org [14:22] Address: 188.40.16.205 [14:25] ljaclark, your server does not have proper reverse DNS entry for itself and that's why you see ip address instead of server name. [14:30] Thank you do you know of a guide that can help me as I used the one on the ubuntu forum to setup bind? [14:40] any idea how I can check which PCI Express version my mobo supports? [14:45] get the model number from lshw and check it on their website? [14:46] patdk-lap: couldn't find anything there - http://www.asrock.com/mb/overview.asp?Model=M3A%20UCC&cat=Specifications [14:46] just says PCI Ex [14:47] AMD 480X [14:49] pcie 1.1 [14:49] thanks [14:49] annoying when they dont say [14:49] then you have to check the northbridge chipset against it's manual :) [16:47] anyone know why I get Hash sum mismatch on a local apt-mirror? [16:49] probably a misconfigured proxy [16:49] or a problem with that local mirror [16:50] qman__, yea I see a forum post about a wrong security mirror [16:51] ah I know....I had to restore the box...and prob had a different mirror.list [16:57] last night i spent 5 hours trying to install ubuntu-server 11.10 and ended up failing [16:58] right now i am at the installing base system phase on a new start today [16:59] what was the issue you ran into? [16:59] it failed at the software selection step every time and also was unable to install grub [17:00] i have pdfs of all the log files [17:01] i wish they were txt files but i had to pull them up on my phone [17:01] so the only way to save them was pdf [17:07] for the software selection bit [17:07] try installing without a network connection [17:07] sometimes that messes things up [17:07] for grub, what is your hard drive configuration? [17:09] i was using 1gb ext4 for /boot , 20gb btrfs / , and 3.7tb lvm ext4 /home [17:10] 3.7tb? what's the physical setup? mdadm, hardware, or fakeraid [17:10] btrfs could be a problem too, I don't know enough about it [17:11] 1tb sdb1 + 3tb sda1 + 125gb sdc2 [17:12] with lvm [17:12] that's really complex, and probably the reason grub isn't working [17:12] it has to be able to determine where the system boots to, and install there [17:13] grub should only care about the /boot partition right? [17:13] no [17:13] grub cares about all of your hard drives and partitions [17:13] it has to determine which drive the bios boots, then install the boot record there [17:13] and then point to /boot from it [17:14] hmm [17:14] http://www.filedropper.com/logs_5 [17:17] i3luefire: using a pastebin is a bit easier :P [17:17] !pastebin [17:17] For posting multi-line texts into the channel, please use http://paste.ubuntu.com | To post !screenshots use http://imagebin.org/?page=add | !pastebinit to paste directly from command line | Make sure you give us the URL for your paste - see also the channel topic. [17:18] RoyK: those pdfs were made last night when i was in ambien land and i thought the only way to get the log files was on my phone [17:19] but thanks [17:19] in any case [17:19] while grub supports /boot on lvm, I don't do it, because it's still far from perfect [17:19] I also disconnect any drives that are not used during the install, and try to make things as simple as is possible for my application [17:19] the boot is not lvm [17:20] the boot is a regular partition [17:20] 1gb [17:20] 1gb for boot? [17:20] and I don't know about btrfs and lvm [17:20] 1gb for boot [17:20] try smaller [17:20] better too big than too small [17:20] no [17:21] ubuntu is not intelligent about kernel updates and simply doesn't remove old kernels [17:21] and if /boot fills up you run into problems installing software [17:21] I usually do 512MB, but 1GB is fine [17:21] qman__, same... [17:23] it looks like from the logs that grub-pc failed to install or something [17:24] I would try ext4 for this partition and see if it works 20gb btrfs / [17:24] who would have thought that a version of bsd (freeNAS) would be easier to install than a version of linux (ubuntu) [17:25] i did ext3 on boot this time and ext4 for root [17:25] same problem? [17:25] boot and root and swap are all simple partitions too [17:25] it is now installing base again [17:26] and the network cable is unplugged [17:26] well, ubuntu is on the so-called 'bleeding edge' [17:26] lol [17:26] pushes new software in spite of the problems [17:26] i guess [17:27] if you want a stable, always easy experience, go with LTS releases or something even more stagnant [17:27] yea I only use LTS for prod [17:27] and wait until .1 [17:29] well my wonderful friend who suggested i use ubuntu server for the easy pkg management of it decided to abandon me during the install last night and he is the one who suggested the 11.10 [17:29] unless you need some of the new features, I'd use 10.04 [17:30] i3luefire, what qman__ said :) [17:30] all my servers are 10.04, except the one still on 8.04 [17:30] should i tell it to install security updates automaticlly [17:30] I don't [17:30] I do [17:30] personal preference [17:30] this is a home server that will be on the net at all times [17:30] ok [17:30] i am enabling it then [17:31] i have made it passed the places where it failed last night [17:32] http://i.imgur.com/thEKg.jpg [17:32] but i changed too many variables so im going back to the start with the fs choice as they were last night and the network unplugged [17:33] the deal with the network is, it tries to update during the install on the fly, so you don't have to later [17:33] but sometimes that messes up, and then apt is broken [17:33] and it can't properly install [17:34] that may have been the entire problem [17:34] judging by the log files [17:35] we shall see [17:39] my friend says "rtfm" so i say to him "there isnt a manual on 11.10 server yet and you are the one who suggested it so you are the fu**ing manual" [17:40] heh [17:40] https://help.ubuntu.com/11.10/serverguide/C/ FYI, but yeah [17:40] i3luefire, https://help.ubuntu.com/11.10/serverguide/C/index.html [17:40] hah [17:40] it's still early, it's probably not complete [17:41] RoyK, wtf is that cable? [17:41] i blame google for my lack of knowledge. lol [17:41] when we switched to upstart the server guide still had tons of incorrect references to /etc/init.d [17:41] thesheff17, it's a euro AC outlet to cat5 [17:41] haha...that is what I thought.... [17:42] does that even work? [17:42] it blows up whatever's on the cat5 end [17:42] lol [17:42] not how you do power over ethernet [17:42] thanks for the manual link tho guys [17:44] :) [17:44] 230VoE [17:47] RoyK: yay, ethernet killer. :) [17:48] fail [17:48] ok. so it must be the partition setup then [17:48] can you purchase that? I would love to send that to some old bosses. [17:48] "here's some extra power for your network" [17:48] I would need 110VoE [17:49] would this do? http://pixdaus.com/pics/HfxKm41AwpIFQRbFXK.jpg [17:50] yes [17:50] i3luefire, I was just reading about btrfs and there is a conversion tool from ext3/4 to btrfs may break your system though [17:52] should i leave the boot at ext3 too? [17:52] instead of ext4 [17:54] shouldn't matter [17:54] I've done both [17:54] I'm almost positive it's the btrfs /, or the LVM that grub is choking on [17:55] well its not the lvm because i left that the same [17:56] btrfs is definitely still unstable, experimental software [17:56] i3luefire: I usually use ext2 on /boot [17:56] no need for a journal there [17:56] hmm [17:56] some distros even mount /boot ro by default [17:57] it would make sense now with grub2, since you don't manually edit anything there anymore [17:57] well i put it on ext3 bc that is what worked earlier and i had it selected already [17:57] remount rw and remount ro could be scripted into update-grub [17:57] IIRC I saw that on SuSE some 5+ years ago, and by then it was a PITA [17:58] back with gentoo 2004, the guide said to simply not mount /boot by default [17:58] grub2 is supposed to support btrfs too [17:58] even on the /boot [17:59] but it cant even do it on the root aparently [17:59] i3luefire: but btrfs is still missing fsck :P [17:59] tru [17:59] something that's been 'in the works' for >2 years [17:59] well, being able to boot btrfs, and being able to figure out and set it up in the installer are two different things [17:59] you don't really need fsck though [18:00] lenios: when the shit hits the fan, you do [18:00] qman__: tru [18:01] ubuntu jumped on ext4 by default a bit early too [18:01] a couple data mangling bugs popped up post release [18:02] new filesystems are just inherently risky [18:02] qman__: ubuntu should be a bit more conservative imho [18:02] but they are so cool tho. lol if they worked... and i guess they do if you know enough about them [18:03] well, it's two goals of ubuntu at odds [18:03] I do have some prod servers with ReiserFS [18:03] making it easy for everyone, and providing the latest and greatest [18:04] those 2 goals are at odds for sure [18:04] good thing im not trying zfs or something [18:04] it chomps thru the ram like crazy from my experience with freeNAS [18:05] I agree that they're just a little too ambitious [18:05] too many show stopping bugs are in release [18:06] yep the release defaults should work 95% of the time not 60% of the time [18:06] that's why I stick to LTS and wait for .1 when it's important [18:06] thats why i dont like fedora [18:07] by default it fails on my desktop hardware. ati opensourse drivers are not set up right or something [18:08] desktops are different, though [18:08] new features are more important than in servers [18:08] my desktop has ati onboard graphics.(no card) [18:08] and hdmi output [18:09] so does mine, still dealing with a number of bugs [18:09] OS drivers have graphical glitches and sometimes hard lock the system, proprietary don't work at all [18:10] isntalling grub failed again at a different step [18:11] but i recovered it [18:11] it just needed me to tell it which drives mbr was available [18:11] yeah, that's been a problem since forever [18:12] even with grub1, as long as I can remember [18:12] once you get more than a couple disks and throw some LVM or raid in, it gets confused [18:12] ok looks like it is booting [18:13] cool [18:13] yea I'm trying 512MB /boot/, 1GB swap, 7GB btrfs / [18:13] w/ ubuntu 11.10 [18:13] server login prompt [18:15] thanks everyone [18:17] do you really need 1GB swap? [18:17] prob not [18:17] the only features i want from btrfs are cow and lzma compression [18:17] I try never to use swap [18:17] i usually use not more than 128M [18:17] it is just a virtual machine [18:18] yea old habbits of 2*ram size [18:18] i have a 6gb swap for my 4gb ram machine [18:18] my desktops generally have no swap [18:18] servers have a little just because [18:18] but it's really not needed if you put in enough RAM [18:19] instead of caching my RAM on disk, I cache my disk in RAM [18:22] i still dont understand why most oses dont let you load the entire os files to ram on systems with 8-16gb ram [18:22] is there an easy way I can get a script to just run once the first time the virtual machine boots? [18:22] /etc/rc.local [18:22] make the script disable itself as the last step [18:22] yep [18:23] i have no idea how to do that tho [18:23] lol [18:23] cool [18:23] very easy with python [18:24] make a script, add it to rc.local, make the script edit rc.local at the end [18:24] then you don't run into self-manipulation issues [18:24] qman__: yeah [18:24] that sounds like a good solution [18:24] says in cat /etc/rc.local that I can just do chmod +x chmod -x and it will start/stop [18:24] not cat... [18:25] man [18:27] no its time to setup transmission, sab, upnp... [18:27] *now its time [18:28] howdy al, i am wanting to migrate the OS (Lucid) of a physical file-server with soft-raid /dev/md0 /dev/md1 etc to a VM on another box -- i was thinking about using Clonezilla but from memory last time i tried to clone partitions of a soft raid i ran into problems ... any advice ? [18:29] instead of cloning partitions, just copy files [18:29] then edit fstab and install grub [18:33] linux is reasonably hardware-agnostic like that [18:33] as long as it can use all the hardware, and knows where its kernel and partitions are, it's good [18:35] probably want to rm /etc/udev/rules.d/70-persistent-net while you're at it [18:54] qman__: ...thanks.. so i would need to remove all the mdadm.conf stuff as well change the UUID's in fstab etc ? [18:54] qman__: IIRC I had an issue with grub last time I tried to duplicate a setup [18:55] pukeko: yes, and the UUID(s) in the grub setup [18:55] pukeko: mdadm.conf can be generated quite easily [18:56] pukeko: mdadm --detail --scan [18:56] Royk: i'm going from soft raid to a VM on a hard-raid [18:56] ok, no need for mdadm, then ... [19:15] yea my ubuntu 11.10 using a root file system of btrfs failed [19:19] thesheff17: did anyone tell you btrfs is experimental? [19:19] yea I was just testing to see if it failed [19:21] worked for me last time I tested it... [19:21] but then, its tools make it rather crappy when being used to zfs :P [19:22] RoyK: did you succeed with raid6->raid5 ? [19:23] yea I tested it on a virtual machine...it was complaining about unmet dependencies with the linux 3.0 kernel [19:25] guntbert: it was just a question - I haven't made the move to raid6 yet [19:25] RoyK: I see - just curious myself :) [19:25] :) [19:27] btw, anyone that knows if it's possible to upgrade the metadata version after creating an md device? [19:28] I have v0.9, being the default, on my home sever, and I see now it doesn't support >2TB drives [19:32] don't think so, to some extent [19:32] I think the metadata locations changed too much between some versions [19:50] patdk-lap: they have, according to the manual, but still, having to recreate your average 10TB md device to use 3TB drives seems a waste [19:51] patdk-lap: and with linux md being one of the more flexible raid solutions out there, adding metadata upgrades would be another fine thing... [19:59] just be aware that raid 6 is _really_ slow [19:59] was saturating gigabit with raid 5, struggling to do 35MB/s with raid 6, and I have pretty good hardware [20:01] wtf? [20:01] how's the CPU load? [20:03] qman__: there's usually a bottleneck somewhere - either that or latency somewhere [20:04] qman, what cpu? [20:12] how would I go about writing a script to move files one by one from one partition to a temp drive, then back to another directory while preserving the structure, etc? [20:14] jeeves_moss: can't you just rsync the bunch? [20:15] RoyK, the "issue" is that I only have a 160Gb "temp" drive. I forgot to turn dedup on when I set up my ZFS drives, so I have to physically move the data off the cluster, then back again. So, I can't just copy everything to another drive, and back. The only way I can take advantage of the dedup is to move the data off of the cluster drives, then back [20:16] jeeves_moss: thank gods you didn't turn on dedup on ZFS [20:16] it's NOT stable [20:16] ?? [20:17] jeeves_moss: I even have a dedicated test machine with some 10TB storage, some SSDs for caching etc, and zfs dedup sucks rather badly [20:17] it is on now. I wanted to run dedup on the data on there right now. hence the request for ideas! [20:17] jeeves_moss: see the mailing lists for more ino [20:17] jeeves_moss: see the mailing lists for more info [20:17] ahhhh. I know it's a new FS, so, I'm testing it on a RAIDZ2 pool. [20:18] jeeves_moss: how much memory/l2arc do you have on the machine? [20:18] jeeves_moss: and what is the pool size? [20:20] RoyK, it's got a gb of RAM (this was a temp soulition untill I get my external deticated NAS box), and the pool is 4.8Tb [20:20] jeeves_moss: then DO NOT USE DEDUP [20:20] jeeves_moss: beleive me on this [20:21] RoyK, is it just not the best preformance, or does it eat your data? [20:21] jeeves_moss: if you have 2GB RAM per terabyte stored and mostly 128kB records, it might work somewhat ok [20:21] jeeves_moss: it'll be dead slow and removing a deduped dataset may take hours or even days [20:22] RoyK, the "plan" is to build this into a 2U box with 8 2.6Ghz cores, and 32Gb of RAM, then exapnd out to ~32Tb of storage [20:22] RoyK, lol, I don't care if it takes days. it's taken almost a month to move all the data off of other boxes! [20:22] I had a server with dedup at a point and removing a dataset took three days, and during that time, it wasn't available for other use [20:23] jeeves_moss: no, it's like it blocks other services [20:23] RoyK, how did you dedup the data though? [20:24] RoyK, I just enabled it, so I was thinking of moving files one by one off to a temp drive, then back again [20:24] jeeves_moss: beleive me - I've done a LOT of testing of dedup, I've read other tests done by people in the opensolaris/openindiana society, and it's not stable, not for production, hardly for testing [20:25] lol, fun! [20:25] jeeves_moss: beleive me on this - I've spent a long time on this, and ended up buying more drives instead [20:26] heh, nothing I have ever done could benifit from dedup [20:26] except work samba shares for their documents [20:26] but that is small enough it doesn't matter [20:26] ahhh, well, the end "goal" of this unit is to be the SAN for a VMWare setup to host a lot of on-line video [20:28] heh? [20:28] san ofr vmware or for videos? [20:28] patdk-lap, if I could "dedup" my wife's bitching, I'd benifit! [20:28] seem two different things [20:28] patdk-lap, it's house both [20:28] heh [20:28] I would do something like I am then [20:29] jeeves_moss: just forget about dedup on zfs for now [20:29] run iscsi/fc/srp/... for vmware using enough space as you need for the vm's [20:29] and just nfs share the videos [20:29] jeeves_moss: or use it if you want to be a crash test dummy of sorts [20:29] could even do vm's on nfs too, makes is easier to backup and stuff, but alittle more annoying otherwise [20:30] patdk-lap, NFS was the plan for the videos. Haven't decided on the proper way to do the iSCSI [20:30] how many esx servers? [20:30] and do you want failover HA? [20:30] jeeves_moss: you're not doing zfs fuse on this? [20:31] on the NAS box right now. I haven't decided for the production box. [20:32] patdk-lap: do you know anything like failover for oi? [20:32] royk, not really [20:32] luckily I don't have to have extreemly fast failover, so manual is perferred [20:33] but for those wanting seemless, clean, auto, nope, don't know anything [20:35] * RoyK just got a message from icinga about a drive failing its SMART tests, and then got another message from a scientist that everything on that server was dead slow - iostat -xdn showed 99% load on that drive alone ... zpool offline ... [20:38] strange, my oi box is rejecting my ssh connections [20:40] what did you do to the poor box? [20:40] I disconnecting it's IB link :) [20:40] everything is fine on it, except ssh [20:41] iscsi, mysql, nfs, samba, all find [20:41] why would ssh depend on ib? [20:41] who said that? [20:41] oyu asked what I did, I disconnected ib [20:42] hmm, ssh service keeps going into mantance mode [20:42] well, reboot the beast :P [20:42] guess I'm going have to get a screen/keyboard on it, and check logs [20:46] fixed ssh [20:49] patdk-lap: what was it? [20:49] bad sshd_config value [20:50] lol [20:50] so not really related to IB? [20:50] na [20:52] * RoyK can only conclude that a backup of 1,5TB over WAN takes its time even with a 60Mbps link [20:54] yep [20:54] my friend attempted a 4tb backup over wan, on a 10mbit link [20:54] he gave up after 3 days [20:55] :) [20:55] some things, sneakernet is still excellent at, or pidgonnet [20:57] patdk-lap: it's only been running for a day or so, and after I offlined that deadish drive, it's running rather quickly [21:02] what's the recommended filesystem to run on ubuntu these days? ext4 is still teh shit and btrfs is not really ready? or? [21:02] I suppose zfs + ubuntu might not be the way to go either. (I'm considering what to use for file storage, software raid) [21:03] today I got 3 drives in software raid-5 running ext4. [21:04] it's only my home server, but I'd prefer not to loose my wedding photos, etc, so at least some parts of the storage should be "safe". [21:04] for now I handle that by rsyncing the important parts to an usb drive every now and then. to give me at least some extra security. [21:06] is there any other way to be secure? [21:07] all it takes is one lightening strike, and all your disks are dead [21:07] well. yeah, youre right. [21:08] so I should look into remote syncing my wedding photos somewhere else. [21:08] or use two different usb drives and have one of them offline at all times. [21:34] so, any comments on zfs or btrfs? know issues with running any of them on ubuntu? [21:37] air_: i run it fine (btrfs) [21:37] air_: but thats on my laptop [21:44] RoyK, Patrickdk, phenom II x4 CPU ~3.2GHz, CPU load is all iowait, have a PCIe x8 card known to be able to saturate all 8 channels, two on a PCI controller I was using before, and two on the onboard [21:46] air_: zfs fuse is a no go performance-wise, btrfs is a no go safety-wise, ext4 works well, and with regular scrubbing of the md device, it should be safe enough [21:47] qman__: then md must be doing some rather heavy amounts of random i/o [21:48] qman, odd, you shouldn't have any issues with that [21:48] unless your doing random writes, and made a large block size [21:48] patdk-lap: with md-based raid6? [21:48] yep [21:48] he should be able to substain >4GB/sec with his cpu [21:49] wonder if my filesystem is fragmented [21:49] using ext3 [21:49] patdk-lap: you'll need a rather large amount of drives for that... [21:49] and it was >90% before I added the last four disks [21:49] RoyK: thanks. and in this case, what does regular scrubbing mean? :) [21:49] royk, therefor raid6 hsouldn't be his issue :) [21:49] but even before that it was not that great [21:49] oh, ext* with >85% is slow as hell [21:50] air_: echo check >> /sys/block/mdX/md/sync_action [21:50] it takes longer and longer for it to locate a free block [21:50] I have had ext3 wait several minutes to locate a free block before, when I was at 99% [21:50] it's 57% now [21:50] patdk-lap: heh - ever tried zfs with 95% full VDEVs? [21:50] royk, nope, only been to 76% so far :) [21:51] at > 90% full VDEVs, ZFS is dead slow [21:51] when I have some spare cash I'll get more disks and redo it, but until then I can't touch it because I don't have anywhere to back up all my data [21:51] but if you used a 256k stripe size, raid6 would have to read that 256k from each drive, then write out atleast 3 drives [21:51] if you only write 4k :( [21:52] 64k chunk [21:52] that shouldn't be too bad [21:52] qman__: what sort of i/o pattern? [21:52] well, mostly large files just being placed [21:53] but I also run torrentflux, so when that's downloading I'd understand [21:53] but when it's just seeding it doesn't make much sense [21:53] or not running at all [21:54] seeding torrents means large number of random i/o [21:55] ya, not as bad as downloading, but still lots of random reads [22:01] * RoyK wonders when linux will get something like l2arc [22:03] thought it did [22:03] flashcache [22:03] works for both l2arc and zil [22:06] all google can tell me about flashcache is something mysql centric [22:06] well, it was originally designed to maky facebook mysql faster [22:06] future plans have the torrents staged on a separate file system, probably a single disk, to alleviate that problem [22:07] qman, I did that [22:07] that drive couldn't substain >4MB/sec transfers cause of that [22:07] maybe a raid 0 then [22:08] but either way, get them off the main data storage to speed it up [22:08] qman__: or a nice set of striped mirrors on zfs :P [22:08] I also want something with checksums, I've had video files get some corruption a number of times [22:08] disks are so unreliable these days [22:09] heh, checksums don't help unreliable disks [22:09] they can only say, you have an issue [22:09] patdk-lap: it helps you detect the problem and with sufficient redundancy, to fix it [22:09] and disks are suppost to have ecc data, rs code [22:09] I mean in the filesystem, to detect when a file gets written wrong [22:09] but seems mostly useless [22:10] patdk-lap: they have ECC all the way, but with the amount of data available today, those ECC algorithms, or sizes, aren't good enough [22:11] https://wiki.archlinux.org/index.php/Flashcache [22:11] royk, well it more that [22:11] when they went to 4k disks, they used a crapload less ecc data [22:11] and also, those sectors are so small, the ecc data is also getting destroyed too [22:11] the physical size of the sector on disk [22:12] thus, use a filesystem with checksumming :P [22:12] * patdk-lap would gladdly replace checksumming with ecc data :) [22:13] hmm, normal 512bytes sector has 40bytes ecc [22:14] a 4k sector has 100bytes [22:15] wish I had the 4k sector with full 320bytes ecc [22:15] but guess they said, it wasn't needed [22:15] seems flashcache isn't in oneiric [22:16] nope [22:16] perhaps in debian sid :P [22:16] but sid is the kid that tend to break toys....