IRC Logs for #crux-devel Tuesday, 2015-03-31

prologicyeah so you guys must have this thing stripped00:00
prologicand many pools on it?00:00
jaegerone pool00:00
prologicoh one giant pool00:00
prologichaha00:00
prologicsrubbing must take ages?00:00
prologicas well as resilvering?00:00
jaeger~19 hours I think the last one took00:01
prologicoh00:01
prologicthat's nothing00:01
jaegerwe don't push it very hard, we use it for the storage and redundancy rather than IOPs00:02
prologicI think I'm getting the impression that some folk are more or less paranoid than other folk00:02
jaegergot an EMC/Isilon OneFS cluster, too, which is a faster performer but much more expensive00:02
prologicand that I shouldn't worry too much and just go with my original plan00:02
prologicwhat's the worsrt that could happen00:02
jaegerYou should keep backups of anything that's important regardless of your NAS, but don't overthink the NAS itself00:02
prologicI have to throw in a few SSD cache disks in future00:02
prologic*nods*00:03
jaegerIf you were running a big web app or a fortune 500 database or something, sure, it'd be more important. For our use cases, though, not so much00:03
prologicspeaking of backups however00:03
prologicone of the main reasons I've done all this is to bring all our data back to home and protect it00:03
prologicfrom silent data rot, corruption, etc00:03
prologicbut how does one prevent backups from being corrupted without another zfs system?00:04
prologicif you loose your nas and you have to restore from backups00:04
prologicthat bit has bothered me and I haven't found a good solution ot that or good advice from anyone :)00:04
jaegerYou can't guarantee the safety of every bit of data 100% of the time, your NAS and your backup could both fail00:04
jaegerthe chances of that, though, are slim if you're careful00:05
prologicthat is true00:05
jaegerbut it's easy to get from careful into way paranoid00:05
prologic*nods*00:05
jaegerI back up my NAS to an extra USB drive, it's not high tech at all. I just want it in case my NAS box fails for some reason00:05
prologicso a good strategy is probably to backup to external USB disks every now and again00:06
prologicfor that important dataset that we can't afford to loose00:06
prologicPersonal data + Family Photos00:06
prologic*nods*00:06
prologicI suppose it should be good enough right?00:06
jaegerat the least I'd keep an external USB backup. If you're worried about physical damage like a fire in your house or something, use crashplan too or take a copy offsite somewhere00:06
prologicCould one take that backup strategy a little further and get two identical USB drives00:06
prologicand do ZFS RAIDz1 or Mirror across them?00:07
prologicon a different machine ofc00:07
jaegerthat would be overkill in my opinion but the only person who can answer the question of "how important is my data" is you00:07
prologicI actually have crashplan00:07
prologicit's really good00:07
prologicand I quite like it as it's all encrypted client-side00:07
prologicpresumably safe from prying eyes :)00:07
prologichehe00:08
prologic*sigh* lots of things can be construed as "overkill" :)00:08
jaegerit's subjective, of course00:08
prologicone might think a rack cabinet at home is overkill or a 24 port Gbps sw00:08
prologic:)00:08
prologicof course :)00:08
prologicof course I enjoy discussing this with you and seeing your insights because let's face it; no-one else I know (closer to my geographically) has done (or are doing) this00:09
prologicmost people I know are very naive about their personal data :)00:09
jaegerto be completely honest it's likely that a simple ext4 NAS would be fine for me. I like ZFS so I use it, it's another toy00:10
jaegerI would keep backups regardless of the FS on my NAS, that's more important00:10
prologicme too :)00:12
prologictoys are fun00:12
prologichaha00:12
jaegerindeed :)00:12
prologicI should take a few pics of my server room one day00:13
prologicit's looking pretty nice00:13
jaegercool00:43
prologicoh a thought...00:46
prologicis the general idea with zfs to create datasets (new zfs filesystems using the zfs tool) to keep things nice and seaprate?00:46
prologicI think snapshots and senc/recv work on teh basis of zfs datasets right?00:46
jaegeralso if you want different features on different filesystems. you might not bother with compression on your media filesystems, for example, but you would use it on your documents00:47
prologicahh yeah good point00:47
jaegerI have a few, like movies, music, tv, backup, etc.00:47
prologicbecause a lot of these settinsg are per dataset/fs right?00:47
jaegeryeah00:48
prologicright now I have compression on on the root pool00:48
prologicI assume if I: zfs create ...00:48
prologicI can turn it off for a dataset?00:48
prologicyeah yeah that's what I was thinking00:48
prologiccreate datasets specific to their use-case00:48
jaegeryes, but note that it isn't retroactive, compressed data will stay that way until it's rewritten00:48
prologicyes ofc :)00:48
prologicI'm thinking00:48
prologiczfs create movies00:49
prologiczfs set movies compression=off00:49
prologicthat sort of thing00:49
jaegeryou can set some inheritance in the child filesystems or you can set each one manually00:49
prologichaha your dataset names are identical to what I had in mind :000:49
jaegercommon uses :)00:49
prologicyeah I think I'm happy with my defaults on the pool as a whole righ tnow00:49
prologicraidz2, compression=on, atime=off, normalization=formD (OS X clients)00:49
prologiceverything else is at it's defaults00:50
jaegerdefaults are pretty reasonable00:50
jaegerchanging them can bite you in the ass sometimes :) I'm in the middle of fixing one of those bites at work, actually00:50
jaegerwhen the ZFS appliance was set up they picked 128k as the block size for the DICOM data filesystem00:51
jaegerthe problem is the average file size is 46k and there are hundreds of millions of them00:51
jaegerso we're wasting shitloads of space on that filesystem00:51
jaegerI changed the block size to 8k recently and I'm shuffling the data00:51
jaegerit's 10T compressed, though, going to take ages00:51
prologicoh wow00:51
prologichaha00:51
prologicoh yeah I'm using ashift=12 here00:52
prologicso 128k block sizes I guess right?00:52
prologicyeah so you have to move the data00:52
prologicrebuid the pool00:52
prologicthen move it back00:52
jaegerI believe that's the default, you can make it smaller00:52
prologicfun :)00:52
jaegerno rebuild needed00:52
jaegerjust move the data so it gets written to new blocks00:52
jaegerwhat I'm actually doing, though, is storing it in uncompressed zip files00:53
prologicwhat do you use at home?00:53
prologicmy drives support 128k blocks00:53
jaeger55000 files into ~15 zip files is great00:53
prologicI imagine my average isn't going ot be 64k files though00:53
jaegerfor home use most files will be larger than 128k, it's a reasonable default00:53
jaegermp3s, movies, etc.00:53
prologicindeed00:53
prologicso all good there :)00:54
jaegeron that fs at work, though, 46k files getting written into 128k blocks is so bad00:54
jaegermore than 50% wasted instantly00:54
jaegerfortunately it's only one filesystem where that's a problem.00:54
prologicahh yeap00:55
prologicashift is a poolw-die setting though00:55
prologicso it affects all datasets?00:55
jaegerper vdev, I think00:56
jaegerI haven't messed with ashift00:56
jaegerprobably wouldn't hurt to tune it but if I remember right it's immutable so that would also require data shuffling00:57
prologicyeah just read that00:58
prologicyeah00:58
prologicand 128k block sizes00:58
prologicthat's an ashift of 17 if my math is right00:59
prologicashift=12 is 4k block sizes00:59
prologicwhich are most modern disks now00:59
prologicbut report 512b for compatibility00:59
jaegerhttps://github.com/zfsonlinux/zfs/blob/master/cmd/zpool/zpool_vdev.c#L108 might be worth a look, too00:59
prologicso yeah my ashift is 12 4k00:59
jaeger2 of my drives are 512b and 2 are 4k :P01:38
prologichaha01:45
prologichow does that work01:46
prologicI guess you have to use 512b across the pool01:46
jaegerthe raid controller reports them all as 512b so my performance is probably crap01:46
jaegerI plan to replace the 2 seagates eventually, then it won't matter01:46
prologicraid controller eh?01:58
prologicI think my limelight tech pc/104 doens't have any raid controllers01:59
prologicI think they put a HBA card in it01:59
prologicI had a quick look under the hood a few nights ago01:59
prologic:)01:59
prologiclooks nice inside01:59
prologicand BIG fans01:59
prologicit's cooler than my media/mythtv box atm :)01:59
prologicsilverstone case uggh01:59
jaegeryeah, that's likely. that's how I use my RAID controller for ZFS, anyway01:59
prologicsmall fans01:59
prologicfair enough01:59
jaegernot RAID, single-unit mode01:59
prologic*nods*02:00
prologicJBOD?02:00
jaegernot JBOD, though, that's bad02:00
prologicyou should think about getting what I got :)02:00
jaegersingle-disk units02:00
prologicthe hw is really quite nice :)02:00
prologicahh02:00
prologicSDU? :)02:00
jaegerI would just use the onboard controller but I can't with VM passthrough, have to pass the RAID controller02:01
prologicahh02:01
prologicdo you pass some of your disk straight thorugh to the vm(s)?02:01
jaegersince my NAS is a VM02:01
prologicoh02:02
prologici see :)02:02
prologicso hw -> vmware02:02
prologicit runs crux+zfs?02:02
jaegerthe 4 disks are directly connect using that RAID controller passed through the VM02:02
prologicgotcha :)02:02
jaegers/connect/connected/02:02
jaegeryes02:02
prologicahhh02:02
prologicis that typical?02:02
prologicI'm going the other way around I guess02:02
jaegerI doubt it but I didn't feel like running another box all the time02:02
prologiccrux+zfs host with kvm, qemu and docker02:02
prologicyeah fair enough02:03
jaegermy ESXi server has room for 4 3.5" disks in a hot-swap bay so I use that02:03
prologicyeah yeah02:03
jaegerthe RAID controller is a bit old, though, might need a new one to get 4k block size. I'll have to research that02:03
jaegerI would just get some kind of SATA/SAS HBA anyway, don't need the RAID bits02:03
prologicspeaking of toys (hard to find) what I would love next is a couple of Intel Atom or similar low-lower compute nodes to go in my rack ideally with 4x 2.5" hot-swap bays -- but with the cavest that the depth can't exceed 19" (550mm)02:04
prologicmy cabinet is shallow :)02:04
jaegeratom and compute node don't really go together, heh02:09
prologichttps://gist.github.com/prologic/a269be0c9216d269cbbe02:13
prologicno they don't - it's not for the raw cmpute02:13
prologicbut for playing with distributed computing more and having more places to put vms/containers/etc :)02:13
prologicdata/media/homes or data/media/users ? :)02:14
prologicfor synced /home mounts over nfs02:14
jaegerup to you, heh02:14
prologicheh02:16
prologicactually data/homes makes more sense02:17
prologicthey're home directories :)02:17
prologicweee02:21
prologicrsyncing down 1/3 servers02:21
prologicthen I can shutdown my codero server02:21
prologicthat thing is costly now because the AUD suck compared to USD right now02:22
prologic$0.76 AUD = $1.00 USD as of this morning02:22
jaegerthe dollar is strong now02:23
prologicyeah your dollar :)02:30
prologicscuky for us though02:30
prologicour tech industry sucks really02:30
prologicwe don't make anything good :)02:30
prologicwee02:35
prologic3 tmux sessions02:35
prologic3 rsync(s)02:35
*** mavrick61 has quit IRC02:38
prologic$ iotop02:38
prologicCould not run iotop as some of the requirements are not met:02:38
prologic- Linux >= 2.6.20 with02:38
prologic  - I/O accounting support (CONFIG_TASKSTATS, CONFIG_TASK_DELAY_ACCT, CONFIG_TASK_IO_ACCOUNTING)02:38
prologicbah :)02:38
prologicnote to self: add a few more kernel options02:38
prologicvmstat and iostat produces interesting results02:38
*** mavrick61 has joined #crux-devel02:39
prologicno going back now :)02:42
*** greduan has quit IRC04:06
*** nilp has joined #crux-devel05:46
*** sudobaal has joined #crux-devel06:22
*** nilp has quit IRC06:32
*** nilp has joined #crux-devel06:37
Romsterjaeger, i dunno how this is gonna work all the LFS and other sites have many rules... i don't know why.07:32
frinnstholy backlog batman!07:40
frinnstgeez, 10 disk for home use?07:42
frinnstand you worry about performance :D07:42
frinnsti run 4 wd green's in raid1007:43
frinnstsoon 4 years old..07:45
frinnstand disks are still more expensive from what they were back then07:46
Romsterand still as slow as back then07:49
prologichaha08:55
prologicyou should see my network graph08:55
prologicoh wait you can08:55
prologichttp://i.imgur.com/zkkUdWE.png08:56
prologicfrinnst, I'm not the one worried :)08:57
prologichaha08:57
prologicothers are just a bit paranoid or pedantic :)08:57
prologicI think RAIDz2 will be fine :)08:57
frinnstis that you wan?08:58
frinnstadorable :)08:58
frinnsthttp://obra.se/mrtg/08:58
prologicyeah09:01
prologiclol09:01
prologicaddorable09:01
prologicyeah yeah don't make fun09:01
prologicAU internet (consumer grade) sucks ass09:01
frinnst100/100 is lovely09:01
prologicmy DSL link is like ~6Mbps/1Mbps09:01
frinnst:(09:01
prologicand off a rim node09:01
prologicso it's shared :(09:01
prologicgeez09:01
prologicso jealous :)09:01
prologicand we Aussies pay for traffic too09:02
prologicper MB09:02
frinnstnow that is fucked up09:02
frinnstwe pay for cellphone data per gb here in sweden09:02
frinnstthat sucks too09:02
prologicanyway been sucking down data off my 3 servers09:03
prologicso let's see if any of them are finisehd :)09:03
prologicyeah most aussie isps chrage per MB09:03
prologicor have traffic quotes09:04
prologicthen charge per MB09:04
prologicvery few have unmetered plans09:04
frinnsti can get 1gbit/100mbit for 75 AUD per month for the first year09:04
prologicit's even worse ont he new Fibre (NBN) network and plans09:04
prologiccause you can chew the 100GB quote in a matter of hours09:04
prologicstop it :)09:04
frinnstthen 136.74AUD per month after that09:04
prologicyou're making me want to move to Sweeden :)09:04
prologic$ df -h09:05
prologicdata/backups/servers   21T   19G   21T   1% /mnt/data/backups/servers09:05
prologicI guess we're not done :)09:05
prologicI'm expecting 60GB+09:05
frinnsti currently pay 71.33 AUD for 100/10009:05
prologicoh man09:05
prologicreally?09:05
prologicgod this country sucks09:05
frinnsti can get 250/100 for 76AUD :D09:06
prologicI'm paying $70 AUD/month for DSL ~6Mbps/1Mbps 100GB ($1/GB after the quota)09:06
frinnstbut 100/100 is enough09:06
crash_i can only get 24/3 here where i live09:06
frinnstthat is fucked up09:06
frinnstadsl crash?09:06
prologicyeah DSL in general sucks in AU09:07
prologicreally really sucks09:07
prologicmost folk are off what we call RIM Ndoes09:07
crash_yeah adsl :( fiber has not been built here in my small town outside of Borås.09:07
prologicbasically Telstra equipment to save cost09:07
prologicthe RIM Node is supplied with some shitty ass link09:07
frinnsti've been with the same isp since 200109:07
frinnstthen it was 10mbit via fiber09:07
prologicthen that's shared amongst like ~1000 or more folk around it09:07
frinnstmade sure i moved to a new apartment with the same fiber connection back in -09 :)09:07
crash_nice09:08
prologicnice09:08
prologicsoo jealous :)09:08
prologicreally09:08
frinnstcrash_: you should move to v�stervik. we are building fiber in the entire county09:08
frinnst:)09:08
crash_ooo09:08
crash_:)09:08
prologicoh cool09:09
prologic2/3 serves are backedup09:09
frinnst2020 ska 90% ha tillg�ng till bredband i v�rldsklass.09:09
frinnstis that you new array prologic ?09:09
crash_frinnst: i wonder if that is true?09:10
prologicoooooh09:10
prologicwait!09:10
frinnstcrash_: they are building everywhere09:10
prologiczfs compression09:10
prologichahahaha09:10
prologicI'm thinking09:10
prologic17G + 8G != 19GB09:10
prologicgeez09:10
prologicfrinnst, yeah :)09:10
prologicjust one dataset09:10
prologicbackups/servers09:10
crash_zfsonlinux? or zfs on bsd?09:11
prologiclinux09:12
prologicon crux to be specific09:12
prologicI've shoved ports in contrib for others to use09:12
prologicalthough I had to upgrade to git HEAD09:13
prologicso maybe I'll make -git versions of the same ports09:13
prologicsome issues with my hw/setup causing race conditions where it wouldn't pickup the new partisions09:13
crash_a zfs port in contrib that is nice :) so there will be 2 ports stable and git version?09:23
crash_you are using ECC memory on your servers?09:25
prologicyeah09:33
prologicI'll push the -git versions up shortly09:33
prologiccrash_, yes09:33
prologicZFS without ECC is kinda pointless IHMO09:33
prologicalthough I'm sure the checksumming algos can probably cope with non-ECC RAM09:34
prologicdunno09:34
crash_yeah that is what i thought :P09:34
prologicjaeger? :)09:34
prologichttp://lime-technology.com/products/avs-10-4/09:34
crash_not fun to do a scrub and you have a bad non-ecc stick :P09:34
prologicthis is what I have at home09:34
*** heroux has quit IRC09:51
frinnstare you using seagate disks?10:31
crash_i have 2 3tb seagate in raid1 and i think they will die soon, as all the other seagate 3tb does?10:34
prologicfrinnst, me?11:55
prologicno11:55
prologicWD/HGST 3TB NAS11:56
jaegerRomster: I have no idea why they use explicit rules but I do know our current docbook-xml port doesn't work properly12:55
jaegerand it was a pain in the ass back when I used it in gnome, even12:56
jaegerprologic: non-ECC is fine until, as crash_ says, you have bad RAM. I'd rather use ECC from the start12:56
prologicagreed :)12:58
prologicI’m not sure ZFS could cope with bad and/or non-ECC RAM really?12:58
jaegerYou might be surprised, it's pretty robust12:59
frinnstnon-ecc is (mostly) just fine12:59
frinnstthe horror scenario would be during a scrub to come across a flipped bit in ram i guess13:00
jaegerprologic: which HBA did they use in it?13:06
prologicAhh I have a link to that somewhere I think13:06
prologiclemme find it13:06
prologichttp://www.newegg.com/Product/Product.aspx?Item=N82E1681318282113:09
prologicactually it’s all onboard13:09
prologicthe damn board has 14 SATA ports!13:09
prologicgeez13:09
prologic:)13:09
jaegerLSI 2308, then13:11
prologicyeap13:13
prologicthat’s the one13:14
prologiccomes up as LSI and MPT Fusion13:14
Romsteri tried your changes jaeger and i don't get the same reuslt on my system.13:26
Romsterresult*13:26
jaegerwhat do you get?13:28
Romsterhttps://gist.github.com/anonymous/2f455d9f547a9cdad8ec13:28
Romsterand i even removed /etc/xml and ran it again.13:29
Romsteralso checked with rejmerge13:29
jaegerwhat does /etc/xml/catalog look like?13:32
Romsterdoesn't exist13:32
jaegerI think you must be using contrib/docbook-xml instead based on that gist13:32
Romsteri copied your Pkgfile  to contrib/docbook-xml then did pkgmk13:33
Romsterand installed it13:33
Romster~/var/cwd/remote/projects/crux/contrib/docbook-xml13:34
Romster$ git diff . |wgetpaste13:34
RomsterYour paste can be seen here: https://gist.github.com/893ed848816fe4da973a13:34
jaegerthat's what the file looks like before my changes13:34
jaegergotta go fix a computer, back in a bit13:34
*** greduan has joined #crux-devel13:35
Romsteri'm heading to bed il;l read what you say later. meantime i found more details on docbook13:35
jaegerback now. Anyway, what I was saying is your gist looks exactly like the file does before my changes. It shouldn't look that way after. My updated version doesn't even run the command that generates this line:14:01
jaeger<rewriteURI uriStartString="http://www.oasis-open.org/docbook/xml/4.1.2" rewritePrefix="file:///usr/share/xml/docbook/xml-dtd-4.1.2"/>14:01
jaegerthe docbook-xsl stuff gets generated either way14:01
jaegerI wonder if we shouldn't just hardcode the rewrite entries because even after my changes it doesn't do *exactly* what you'd expect. It does work for my tests so far but it's still weird.14:31
jaegerprologic: think I might replace my RAID controller with an IBM M1015 and flash its firmware to remove the option ROM15:03
jaegerI should do some benchmarks with my current controller as well15:09
jaegerI'm guessing I'd get better performance (at the cost of a small amount of storage space) by switching to ashift 12 but my controller is only sata 2 anyway :)15:09
jaegeragain nothing that REALLY matters for home use but I do like to play with the toys. :)15:10
jaegerRomster: re-reading my above comments I was wrong; it SHOULD generate that rewriteURI line but it should also generate 4 others for each version of docbook-xml DTDs. or at least the "rewriteSystem" versions15:12
jaegercolor me still confused15:12
jaegerprologic: ended up ordering an LSI 9240-8i from ebay for a good price ($89 USD)15:40
*** heroux has joined #crux-devel16:48
*** Feksclaus has joined #crux-devel17:13
jaegerports in contrib with perl 5.16 in their footprints: exiftool p5-cairo p5-digest-sha1 p5-extutils-pkgconfig p5-gd p5-glib p5-gtk2 p5-io-string p5-net-ip p5-pango p5-rpc-xml p5-text-iconv p5-unix-syslog spamassassin19:43
jaegerlooks like they're mike_k, sepen, alancio, and teK_ ports19:44
prologicjaeger, nice19:45
jaegerI'll probably shuffle my data off the pool and make one with ashift set to 12, depending on the benchmarks, I guess19:46
jaegerseems like there's no real reason NOT to use 12 instead of 9 when you have 4k disks, even if you have some 512b ones in the mix19:46
frinnstiirc alan is still using 2.8 or something ancient19:47
frinnstand lazy tek is still on 3.019:47
frinnst:p19:47
jaegermaybe he can skip 3.1 altogether, hehe19:48
frinnstalso prepare your cpu coolers! firefox 37 will soon be pushed19:48
frinnstmuhahah19:48
jaegerbring it!19:48
jaegereven some perl 5.12 in footprints19:51
frinnstoh wow19:51
frinnstalan?19:51
jaegerthe one I looked at is moritz19:52
frinnstis he still active?19:52
frinnstiirc he just dropped of the earth more or less19:52
jaegerI'm guessing not but not sure. Maybe we should do some git stats monitoring19:52
frinnstsame with rugek19:53
frinnsti sent him a couple of emails a few years ago. never heard anything back19:53
frinnstthese days the emails bounce19:53
frinnsthope he is ok19:53
jaegerprobably just didn't have time anymore19:53
frinnsthope so19:54
jaegerhttp://jaeger.morpheus.net/linux/crux/gs/authors.html20:02
jaegerthat's for contrib, was curious about authors20:02
frinnstcool20:04
frinnstwhat did you use to generate that?20:04
prologicdata/backups/servers   21T   22G   21T   1% /mnt/data/backups/servers20:06
prologicinteresting20:06
prologicso 28.5G compressed down to 22G20:07
prologicall servers sycned I can shut them down now :)20:07
prologicexcept I have to move bind and lighttpd services somewhere else pronto! :)20:07
frinnstI run btrfs in "production" because of compression20:08
frinnstits epic for elasticsearch20:08
prologicdata/backups/servers  available             20.4T                      -20:09
prologicdata/backups/servers  available             20.4T                      -20:09
prologicdata/backups/servers  referenced            21.7G                      -20:09
prologicdata/backups/servers  compressratio         1.40x                      -20:09
prologicyeah20:09
prologiccompression is nice20:09
prologicwhat kinds of datasets would you not set compression on though?20:09
prologicobviously uncompressable files - but what?20:09
frinnstwith btrfs? nothing20:10
frinnstthere is next to no penalty20:10
prologicsame here as I understand it20:11
prologicespecialyl with this Xeon something a rather20:11
prologicIntel(R) Xeon(R) CPU E3-1231 v3 @ 3.40GHz20:12
frinnstno fancy utils to calculate btrfs savings with compression20:13
frinnstTotal devices 2 FS bytes used 2.31TiB20:13
frinnstfredrik@nibbler:~$ du -shx20:14
frinnst2.2T.20:14
frinnstthats the best I got :)20:14
prologicjaeger, http://zfswatcher.damicon.fi/20:14
prologicI'm installing this on my system20:14
prologicand I'll chuck that into contrib too20:14
prologicit's quite nice IHMO20:14
prologichehe20:14
prologicfair enough20:14
prologicyeah funny thing about zfs I guess is that the reported used space is the compressed space20:15
prologicso you have no idea how big the dataset is uncompressed I guess20:15
prologicnot that it matters?20:15
frinnsthttps://btrfs.wiki.kernel.org/index.php/FAQ#How_much_free_space_do_I_have.3F20:18
jaegerhttp://jaeger.morpheus.net/linux/crux/files/firefox%2337.0-1.pkg.tar.xz if anyone wants it20:44
jaegerfrinnst: gitstats, http://gitstats.sourceforge.net/20:44
jaegerthe compression penalty is slight extra CPU usage20:45
jaegernot a big deal20:45
jaegersome data isn't worth even trying to compress, though20:45
jaegerprologic: looks like a decent frontend20:46
prologicyeah I thought so too20:51
prologicjust for monitoring ofc20:51
prologicbut has notifications as well which is nice20:51
prologicwill set this up along with smartd20:52
jaegerwhich compression algorithm did you pick for your pool?20:53
prologiclz4 I think20:56
jaegerthat's probably the best choice for the most part20:57
prologicI believe so20:57
jaegerprologic: I haven't looked at your version of the zfs ports but one thing I'd suggest adding if you didn't already: the /etc/zfs dir21:05
jaegerbecause if /etc/zfs doesn't exist, /etc/zfs/zpool.cache doesn't get created21:06
jaegerI added that to my system but the original zfs port I made didn't have it21:06
prologicofc!21:08
prologicI'll do that right now :)21:08
jaegerzdb requires it to work properly. There are probably other good reasons to have it but that's why I noticed it21:09
prologicyeah yeah you're absolutely right21:12
prologicadding it now21:12
*** sudobaal has quit IRC21:14
prologichmm21:17
prologichttp://www.umart.com.au/umart1/pro/Products-details.phtml?id=10&id2=126&bid=2&sid=219391&21:17
prologicI wanna buy one of these as backup device21:17
prologic:)21:17
jaegerI think some model of microserver is what frinnst uses21:17
prologicas a NAS in general?21:20
prologicyeah I wanna use it as my backup device21:20
prologichaha21:20
prologicI mean what are the chances really of loosing 14 disks and two devices at the same time? :)21:21
prologicthrow in crashplan for the important stuff21:21
prologicand bob's ya uncle :)21:21
frinnstwhat, you cant choose compression on the fly with zfs?21:23
frinnsthow lame :)21:23
frinnstprologic: i have 3 of those :)21:24
frinnstactually just one. the other is n40l and nl3621:24
frinnstthey are brilliant21:24
jaegerfrinnst: you can, I was just curious which he picked21:25
frinnsthttp://obra.se/ikea1.jpg, http://obra.se/ikea2.jpg21:26
prologicjaeger, you're going to say using these as backups is overkill right? :)21:29
prologicremember me paranoid discussion of protecting data and backups from corruption? :)21:29
jaegerin my opinion, sure. I still say it's subjective, though. If you want that, use it21:30
*** mechaniputer has joined #crux-devel21:30
frinnstI use one as pure backup21:31
frinnstthe other one hosts obra.se, dns mail http firewall and all sorts of stuff21:31
jaegerA disadvantage of my use of external USB drives for backups is that they need to be plugged into an existing machine somewhere21:31
frinnstgen8 is so-so21:31
frinnstyou can get it with a xeon cpu21:31
frinnstand the stackable switch is sexy as hell21:31
prologicjaeger, yeah21:32
prologicI've also found using USB as backups to be disadvantegous in rreboots21:32
prologicif you use the UUID of the drive21:32
prologicthe system doesn't boot for some reason21:32
prologicdue to some weird race condition somewhere21:32
jaegerI haven't run into that myself21:33
frinnstzzz21:34

Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!