#jogamp @ irc.freenode.net - 20130529 05:05:22 (UTC)
20130529 05:05:22 -CatOut- Previous @ http://jogamp.org/log/irc/jogamp_20130528050508.html
20130529 05:05:22 -CatOut- This channel is logged @ http://jogamp.org/log/irc/jogamp_20130529050522.html
20130529 05:06:04 <[Mike]> for the most part I (my code) don't see the need to care, however ardor3d's JoglNewtWindow adds a windowDestroyNotify callback and sets the _isClosing property to true.
20130529 05:13:57 <[Mike]> Another thing I've noticed. When reparenting the GLWindow on my secondary monitor, a new GLContext is created. The ardor3d joglcanvasrender uses only the initial context set to it. I can fix it up by manually setting the context later, however I don't know when the GLContext is created.
20130529 05:14:35 <[Mike]> Is there any kinds of hooks for that? I seemed to get around the issue by setting preserveGLStateAtDestroy(true), if that is a valid resolution.
20130529 05:36:09 * [Mike] (~Mike]@anon) Quit ()
20130529 06:23:44 * monsieur_max (~maxime@anon) has joined #jogamp
20130529 09:59:53 * monsieur_max (~maxime@anon) has left #jogamp
20130529 10:56:46 * monsieur_max (~maxime@anon) has joined #jogamp
20130529 12:21:47 <sgothel> our jogamp.org raid1 setup exposes block mismatches .. :(
20130529 12:22:02 <sgothel> I checked this yesterday .. and the semantics may vary .. i.e. unclear
20130529 12:22:42 <sgothel> however, it boils down to the fact that non of the Linux SW raid is transaction safe, i.e. silent corruption may happen.
20130529 12:23:28 <sgothel> this brought me back to re-reading ZFS, which claims to be fail-save w/ raid-z (you know that I guess)
20130529 12:25:05 <sgothel> since may this year, 'zfs on linux' is marked stable by the LLNL guys .. - http://zfsonlinux.org/
20130529 12:25:31 <sgothel> (-> march 2013)
20130529 12:25:50 <sgothel> anybody here has own experiences w/ ZFS already ?
20130529 12:25:57 * rmk0 prods jk4
20130529 12:26:25 <sgothel> [root@jogamp sysconfig]# cat /sys/block/md2/md/mismatch_cnt
20130529 12:26:25 <sgothel> 1024
20130529 12:26:32 <sgothel> a scary value ..
20130529 12:26:35 <rmk0> nasty
20130529 12:26:49 <sgothel> I scrubbed the disks yesterday .. same result :(
20130529 12:27:05 <rmk0> jk4's tried zfs on linux
20130529 12:27:07 <sgothel> here at home, I have similar setup .. all good
20130529 12:27:10 <rmk0> not sure how it went
20130529 12:27:15 <rmk0> i use zfs here on freebsd
20130529 12:27:18 <rmk0> ♥
20130529 12:27:26 <sgothel> but will move to zfs ..
20130529 12:27:29 <sgothel> oh .. great
20130529 12:28:10 <sgothel> so current server .. has 2 HDD .. then I guess a silly raidz-1 w/ 1N+1P should suffice right ?
20130529 12:28:28 <rmk0> would think so
20130529 12:28:55 <sgothel> hmm .. how about the boot / root part ?
20130529 12:29:03 <rmk0> i've not touched that side
20130529 12:29:13 <rmk0> i boot from ufs, but use zfs for storage
20130529 12:29:48 <sgothel> so I guess .. I setup a new machine .. minimal crap .. then rest (/var/* /srv/* /home/* .. etc) on zfs pool
20130529 12:30:04 <rmk0> yep
20130529 12:30:15 <sgothel> "/var" might be tricky ..
20130529 12:30:33 <rmk0> yeah, depends on how the OS boots
20130529 12:30:53 <sgothel> for home file server .. I read that raidz is a VDEV, and more VDEV -> more perf, i.e. IOPS
20130529 12:31:07 <sgothel> so striped raidz-1 might be good ?
20130529 12:31:20 <sgothel> i.e. you have 6 hdd ..
20130529 12:31:40 <sgothel> you could do 1 raidz-2 (4+2) or 2 striped raidz (2+1)
20130529 12:31:41 <rmk0> most i've used is three so far
20130529 12:31:50 <rmk0> that was striped with parity
20130529 12:32:01 <rmk0> was extremely fast
20130529 12:32:06 <sgothel> you mean raidz1 ?
20130529 12:32:16 <rmk0> yeah, think so
20130529 12:32:26 <sgothel> memory ?
20130529 12:32:35 <rmk0> that was on a machine that had 12gb...
20130529 12:33:26 <rmk0> am currently running a backup file server that's using two disks mirrored, with 2gb ram
20130529 12:33:28 <sgothel> we are on 8GB .. all used already for cache :)
20130529 12:33:41 <rmk0> yep, it loves memory
20130529 12:33:47 <sgothel> so I guess 32GB would not hurt
20130529 12:34:03 <rmk0> i think it disables most caching with less than 4gb
20130529 12:34:05 <sgothel> and that ECC memory
20130529 12:34:07 <rmk0> 32gb will be fine
20130529 12:34:36 <sgothel> hetzner.de has 69/79 EUR per month .. offers
20130529 12:34:40 <sgothel> http://www.hetzner.de/hosting/produktmatrix/rootserver-produktmatrix-ex
20130529 12:34:43 <rmk0> the main thing i like about it is the command line tools
20130529 12:34:49 <sgothel> EX6 / EX 6S
20130529 12:35:02 <sgothel> yup .. like 'all in one' :)
20130529 12:35:09 <sgothel> all I miss is loop-aes included :)
20130529 12:35:16 <rmk0> they're extremely "literal" and explanatory
20130529 12:35:18 <sgothel> I have to think about that for home server
20130529 12:35:34 <sgothel> loop-aes -> strong encryption
20130529 12:35:55 <rmk0> yeah
20130529 12:36:01 <sgothel> but using loop device .. well, would take away the 'use whole disk' performance of zfs
20130529 12:36:18 <sgothel> uh well, can't have it all :)
20130529 12:36:29 <sgothel> any better deals around than hetzner ?
20130529 12:36:51 <rmk0> this is more jk4's area
20130529 12:37:01 <rmk0> i've never bought a vps
20130529 12:37:20 <sgothel> 79 EUR: (32GB ECC, 2x 4TB enterprise, Intel Xeon)
20130529 12:37:36 <sgothel> I guess I give them a call ..
20130529 12:37:48 <sgothel> so how to make transition .. ?
20130529 12:38:02 <sgothel> I can prepare new machine .. but at one point .. we have to switch :)
20130529 12:38:15 <sgothel> i.e. all write access must be disabled .. (wiki, email, ..)
20130529 12:38:22 <sgothel> final rsync
20130529 12:38:34 <sgothel> ip4/ip6 switch
20130529 12:38:35 <rmk0> right
20130529 12:38:49 <sgothel> then .. which OS ?
20130529 12:39:11 <sgothel> right now we have CentOS5 .. I prefer Debian7
20130529 12:39:21 <rmk0> yeah, would take debian over centos
20130529 12:39:32 <sgothel> CentOS / RedHat .. I always need to find the right config files :)
20130529 12:39:41 <sgothel> great
20130529 12:39:53 <rmk0> wonder if they have a zfs kernel package
20130529 12:40:09 <sgothel> http://zfsonlinux.org/debian.html
20130529 12:40:13 <sgothel> indeed :)
20130529 12:40:32 <rmk0> phew
20130529 12:40:42 <rmk0> building kernels is a fucking drag
20130529 12:41:05 <sgothel> hope they haven't added too many backdoors in their implementation .. LLNL, DoD, .. hehe
20130529 12:41:14 <rmk0> /o\
20130529 12:41:14 <sgothel> not so much w/ debian .. quite nice
20130529 12:42:48 <sgothel> .. and Oracle is still sitting on the other new ZFS stuff (encryption, etc) .. well, if all fails, in a few years there might be another great FS :)
20130529 12:43:40 <sgothel> right now it seems the zpool is same for Ilumus, FreeBSD, Linux, .. wonder whether they are binary compatible
20130529 12:43:45 <sgothel> storage I mean
20130529 12:43:48 <rmk0> think they are
20130529 12:44:01 <rmk0> the pool format is versioned
20130529 12:44:11 <rmk0> don't think they'd bother with that without cross-OS compatibility
20130529 12:46:56 <sgothel> we have a translation of http://science.webhostinggeeks.com/opengl-evolucija-i-jogl :)
20130529 12:48:20 <sgothel> " Serbo-Croatian language, that is used in all Former Yugoslav Republics"
20130529 12:49:42 <sgothel> so they haven't translated the method names of the little sample, funny :)
20130529 12:52:25 <xranby> hi all
20130529 12:52:33 <sgothel> Hi Xerxes
20130529 12:53:14 <sgothel> my reply regarding invulnerabilities is in my draft folder, will continue later w/ it .. :)
20130529 12:53:50 <sgothel> the raid1 block mismatch opened another door of excitement .. hehe
20130529 12:54:16 <xranby> hi sgothel, ok thank you, personally I like the idea of early full disclosure and document defects in the bugzilla as fast as possible
20130529 12:54:42 <sgothel> me not so much :)
20130529 12:54:57 <sgothel> however .. most of the stuff doesn't concern us
20130529 12:55:40 <sgothel> i.e. the native lib loading code is not exposed to public in a generic form and with elevated privileges
20130529 12:55:50 <xranby> i look forward to the mail
20130529 12:55:59 <sgothel> they were merely pointing to a bug of the JVM impl.
20130529 12:56:53 <sgothel> our only required exercise is to check whether we expose anything critical in a generic form with elevated privileges .. and buggy native commands ofc
20130529 12:57:38 <sgothel> yeah, lets discuss it later after my first review
20130529 13:17:45 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20130529 13:25:19 * monsieur_max (~maxime@anon) has joined #jogamp
20130529 14:01:01 <sgothel> Rami: May 31 Siggraph tickets become more exp. !!
20130529 14:28:04 * monsieur_max (~maxime@anon) Quit (Read error: Connection reset by peer)
20130529 14:53:08 * xranby_ac100 (~xranby@anon) has joined #jogamp
20130529 16:32:52 * xranby_ac100 (~xranby@anon) Quit (Read error: Connection reset by peer)
20130529 16:57:53 * monsieur_max (~maxime@anon) has joined #jogamp
20130529 17:03:48 * xranby_ac100 (~xranby@anon) has joined #jogamp
20130529 17:07:22 * [Mike] (~Mike]@anon) has joined #jogamp
20130529 18:33:09 <sgothel> ordered: http://www.hetzner.de/en/hosting/produkte_rootserver/ex6s (will be installed in 1-2 days, testing stuff locally w/ Debian & ZFS now)
20130529 18:34:39 * DemoscenePassiv (~Lutsche@anon) has joined #jogamp
20130529 20:05:40 <rmk0> nice
20130529 20:06:25 <sgothel> doing a local dummy install now, i.e. install debian on USB stick from rescue system (USB as well)
20130529 20:06:45 <sgothel> then install zfs stuff incl. zfs-initramfs
20130529 20:07:03 <sgothel> well .. then config zfs on actual new pool .. and trying to move all over :)
20130529 20:07:22 <sgothel> the server config allowed to request a 16GB USB stick - savor :)
20130529 20:07:27 <sgothel> savior :)
20130529 20:08:00 <sgothel> I write down everything .. and commit later to git
20130529 20:32:57 <sgothel> so I guess we can use 3 datasets: 'jogamp', 'users' and 'system' w/ intention to have all essential data regarding our project in 'jogamp'
20130529 20:33:08 <sgothel> maybe adding 'jogamp-data' as well ?
20130529 20:33:34 <sgothel> benefit: those datasets are of variable size and can be managed independently .. (backup, ..)
20130529 20:34:06 <sgothel> ofc .. would need to mount them all across the actual rootfs .. but this is possible
20130529 20:36:05 <sgothel> and 'var'
20130529 20:43:57 <rmk0> yeah, they can be nested too, which is nice
20130529 20:44:02 <rmk0> jogamp/jogl, jogamp/joal, etc
20130529 20:44:24 <rmk0> i give each machine that submits backups here a filesystem in /backup
20130529 20:44:39 <rmk0> means space usage can be tracked easily
20130529 20:58:10 <rmk0> i recommend turning on gzip compression too, if it's not already enabled
20130529 20:58:27 * void256 (~void@anon) has joined #jogamp
20130529 21:41:15 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20130529 23:33:47 * DemoscenePassiv (~Lutsche@anon) Quit (Ping timeout: 252 seconds)
20130529 23:37:52 * void256 (~void@anon) Quit (Remote host closed the connection)
20130530 02:43:59 * jk4 (~jk4@anon) Quit (Quit: Lost terminal)
20130530 05:02:18 * [Mike] (~Mike]@anon) Quit ()
20130530 05:05:35 -CatOut- Continue @ http://jogamp.org/log/irc/jogamp_20130530050535.html