#ceph IRC Log


IRC Log for 2011-02-20

Timestamps are in GMT/BST.

[1:09] * tel_ (~tel@59.64-150-148-net.sccoast.net) Quit (Remote host closed the connection)
[1:16] * greglap1 (~Adium@cpe-76-90-239-202.socal.res.rr.com) has joined #ceph
[1:16] * greglap (~Adium@cpe-76-90-239-202.socal.res.rr.com) Quit (Read error: Connection reset by peer)
[2:35] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) has joined #ceph
[4:04] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) Quit (Quit: This computer has gone to sleep)
[4:05] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) has joined #ceph
[4:09] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) Quit ()
[4:14] * votz (~votz@dhcp0020.grt.resnet.group.upenn.edu) Quit (Quit: Leaving)
[10:05] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) has joined #ceph
[10:51] * votz (~votz@dhcp0020.grt.resnet.group.upenn.edu) has joined #ceph
[10:55] * allsystemsarego (~allsystem@ has joined #ceph
[14:26] * pombreda (~Administr@ has joined #ceph
[14:27] <pombreda> Howdy ..
[14:27] <pombreda> fwiw http://discussion.dreamhost.com/thread-128977-post-134228.html
[14:28] <pombreda> this puppy is behaving quite nicely so far I shall say :)
[14:28] * pombreda bows to the ceph dudes
[14:31] <DeHackEd> How much load though? transactions/second? megabites?
[14:31] <DeHackEd> that's the important stuff. how does it hold up to pressure?
[14:39] <pombreda> DeHackEd: well atm only 13 terabytes
[14:40] <pombreda> DeHackEd: the goal is to let know this baby who is his daddy
[14:41] <pombreda> DeHackEd: this is not a formal becnhmark, but rather a practical evaluation based on a real lfe use case
[14:41] <pombreda> on a lota of bytes
[14:42] <pombreda> DeHackEd: my use case in the target range of ~ 2 pbytes
[14:43] <pombreda> with mostly reads, WORM-style archival
[14:44] <pombreda> I am not mesauring throughput for now..just trying to fill up the puppy to the rafters a bit, next will come using it
[14:45] <pombreda> DeHackEd: and i have no idea of the specs of the HW the dreamhost playground runs on
[15:36] <darkfader> "I am using only the file storage for now, I may be a bit old fashioned but I could not care less for S3-like object storage" hehe my words
[15:37] <darkfader> if you store 6 TB and use 12TB then it is probably replicated and ok
[15:39] <darkfader> pombreda: how big are your archival files?
[15:40] <darkfader> im my old job the archive files i.e. had single tiff files for signatures, like 3kb and millions of them :)
[15:40] <pombreda> ~ atm 6TB these are widely varrying in size from small few k code files to large 4GB ISOs
[15:41] <pombreda> when complete this will be 160Millions files for 11TB real content
[15:41] <pombreda> these are open source stuffs available on public mirrors that I rsync
[15:42] <pombreda> darkfader: so this is quite varied
[15:43] <darkfader> i think anything over 4MB is no worry, and isos will be the ideal food. the smaller is interesting. and 160mio files is really like a real world filesystem, thats a good test
[15:43] <pombreda> darkfader: average size might be ~ 70 bytes 11billions bytes/168M files
[15:43] <pombreda> darkfader: but again lots of variations
[15:45] <pombreda> darkfader: this is as real as it gets imho, albeit very software-world centric, these are open source code bits (source and binaries)
[15:45] <pombreda> but at a scale large enough that there is everything you could think of
[15:45] <pombreda> including sparse VM images, etcs and one liner files
[15:45] <darkfader> if you need something good for testing small io pain, use the oracle documentation cd
[15:46] <darkfader> it's full of html docs, the slowest area in my backups :)
[15:46] <pombreda> :D
[15:49] <darkfader> in my first test (last year in april???) i still managed to break the mds with a few different sized rsyncs
[15:49] <darkfader> but i think today you'll have to work hard to break anything
[16:10] <pombreda> darkfader: :) are you a ceph dev?
[16:10] <darkfader> no!
[16:11] <darkfader> interested small-time user
[16:12] <DeHackEd> I like the idea of ceph. from a sysadmin standpoint I might consider its use in place of an iSCSI server (or rather, server cluster) but there's some kinks to get around first
[16:51] * sakib (~sakib@enjoyable.benefit.volia.net) has joined #ceph
[16:51] * sakib (~sakib@enjoyable.benefit.volia.net) Quit (Read error: Connection reset by peer)
[16:52] * Yoric_ (~David@dau94-10-88-189-211-192.fbx.proxad.net) has joined #ceph
[16:52] * sakib (~sakib@enjoyable.benefit.volia.net) has joined #ceph
[16:57] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) Quit (Ping timeout: 480 seconds)
[16:57] * Yoric_ is now known as Yoric
[17:06] * verwilst (~verwilst@dD576FAAE.access.telenet.be) has joined #ceph
[18:19] * verwilst (~verwilst@dD576FAAE.access.telenet.be) Quit (Quit: Ex-Chat)
[18:28] * johnl_ (~johnl@ Quit (Quit: Reconnecting)
[18:29] * johnl (~johnl@johnl.ipq.co) has joined #ceph
[18:30] * sakib (~sakib@enjoyable.benefit.volia.net) has left #ceph
[19:33] * allsystemsarego (~allsystem@ Quit (Ping timeout: 480 seconds)
[19:58] * jantje_ (~jan@paranoid.nl) has joined #ceph
[20:00] * jantje (~jan@paranoid.nl) Quit (Ping timeout: 480 seconds)
[20:11] <wido> anyone around?
[20:12] <wido> I've got a Ceph-unrelated issue. I accidently removed my partition table by doing mkfs.btrfs on my sda...
[20:12] <wido> Machine is still running, but I can't get the exact partition table back the Ubuntu installation created
[20:13] <wido> the values in /sys/block/sda1, sda2 and sda5 should tell me what to do
[20:13] <wido> "start" and "size" should be enough.
[20:13] <wido> Anyone tried this before?
[20:17] <darkfader> wido: weekend fun? :)
[20:18] <wido> darkfader: No, not really :) Was playing around with Ceph and broke my test machine. I really do hope the power doesn't fail
[20:18] <darkfader> what files in /sys/block do you look at?
[20:18] <wido> I've saved the values from /sys
[20:18] <wido> start and size of sda1, sda2 and sda5
[20:18] <wido> sda1 = ext2, just /boot
[20:18] <wido> and sda5 is LVM where the root fs is on
[20:19] <darkfader> this would be in interesting exercise
[20:19] <darkfader> but i just woke from suddenly falling asleep
[20:19] <darkfader> anyway, right now i just dont see where in /sys/device/sda you found partition entries
[20:20] <darkfader> bdi dev holders queue removable slaves subsystem
[20:20] <darkfader> capability device power range size stat uevent
[20:20] <darkfader> i got those for xvda
[20:20] <darkfader> oh, and i should switch off the mail/web server when following you
[20:20] <darkfader> really the wrong box to test on
[20:21] <darkfader> i found a useless box for testing.
[20:22] <wido> darkfader: The "start
[20:22] <wido> and size values should give you a clue of the sectors
[20:22] <wido> which to use
[20:22] <wido> darkfader: http://serverfault.com/questions/146892/repair-partition-table
[20:23] <darkfader> i dont got the files in my old centos kernel it seems
[20:23] <wido> But fdisk refused me to created partitions with that values, but "parted" just did
[20:23] <darkfader> oh
[20:24] <wido> I'll see if it works when I reboot the box tomorrow morning :)
[20:24] <darkfader> did you back it up yet?
[20:24] <darkfader> also, you can just partprobe to find out about that
[20:24] <wido> Yes, pulled my data of it. But really do not want to re-install
[20:24] <darkfader> avoid the reboot
[20:25] <wido> I tried testdisk, that one saw my partitions inside LVM
[20:25] <wido> darkfader: I'd like to know if the machine still reboots, otherwise I have to re-install anyway
[20:25] <darkfader> unix is when you can fix it.
[20:25] <darkfader> ;p
[20:25] <darkfader> lvm should always use live info, so try a pvscan -vvv for a test
[20:26] <darkfader> (you should have caching disabled in /etc/lvm/lvm.conf and no /etc/lvm/cache/.cache around)
[20:28] <darkfader> next time do that on hpux and i'll fix it for free
[20:28] <darkfader> ;)
[20:30] <wido> Yes, I'll try
[20:30] <wido> tnx :)
[20:59] * alexxy[home] (~alexxy@ has joined #ceph
[20:59] * alexxy (~alexxy@ Quit (Ping timeout: 480 seconds)
[21:29] * pombreda (~Administr@ Quit (Quit: Leaving.)
[23:54] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) Quit (Quit: Yoric)

These logs were automatically created by CephLogBot on irc.oftc.net using the Java IRC LogBot.