mercutio: SMP memtest86 doesn't work particularly well in my experience. It's liable to lock up randomly
plett: oh wow
not good when peopel use it to test stability
Well, I use it to test RAM ;)
That's normally running on a new server overnight while the disks are in another machine having badblocks run on them
do you ever find memory problems?
i've found memory problems on a desktop and it was really quick.
Once in a while, yes
was only odd or even addresses
can't remember which, buut it was obviouuusly one of the sticks
so i guessed which one it was, and it got way worse
so i knew it was the other one that was good :)
A more likely scenario is that a server will start logging ECC errors during normal operation and we'll take it out of service and PXE boot memtest on it
but with ecc i dunno if it'd show issues as easily
does it pick up on ecc errors?
ahh ok
Yep
I also once span up a load of virtual machines running memtest to see how responsive other guests would be under high cpu+memory load
how'd that go?
It worked fine, I couldn't even notice the extra load from inside the other guests
cool
i wonder if it'd be worse if it seeked randomly around memory
Not like running multiple bonnie++ tests to simulate heavy disk IO, that's what will really cause noticable problems
yeh disk i/o is one of the biggest issues with virtualisation
that's why even cheap vps places have often gone to ssd
i know xen has some disk scaling issues too
i dunno about kvm
but 30k iops on a vm wouldn't be fun
really need a combination of hard-disk/ssd
but all that stuff is kind of premature
Indeed. ZFS does combined hdd+ssd well with it's ZIL and L2ARC
But I can't switch my virtual machine hosts to FreeBSD to get ZFS until bhyve is much more supported by other tools
i'm doing zfs on linux
the problem is that on reboot it loses the whole read cache
ime it doesn't seem to help a lot more than having a largish amount of ram spare
I wouldn't use zfs-on-linux on a production VM host though
why?
Lack of experience really, and I don't see it being a longterm solution so don't want to put the time into getting the experience
ahh ok
it's pretty stable.
i used to do openindiana/zfs/xen
(When ECC errors occur, the memory does report the error and that shows in Memtest, as well as OS logging depending upon configuration)
If it can't ever be merged into the mainline linux kernel, it's only ever going to be a niche thing, and once btrfs gets good enough to handle most workloads, the niche will get smaller and smaller
brycec: i just didn't knwo if memtest supported that.
plett: it's a huge niche thing though
"once btrfs gets good enough"
btrfs is making surpisingly slow progress towards stability.
Yeah, I know. They're taking their time about it :)
slow enough with enough developers that makes me wonder about the underlying systems.
i dunno why but i've had multiple incidents with fs corruption when i've tried btrfs.
hmm, ext4 can have an extenral journal can't it?
i wonder if having a journal for ext4 on ssd raid would work well
I'm only using btrfs on non-critical boxes like backups of backups
I don't know about external journals, I've never looked in to it
i was only using it on non-critical systems too
and it gave me enough headaches to make mwe wonder.
actually that was made worse by using ssd's
as i had small /
and it's easy to run out of space with lots of files by default
but besides that, even with ssd there'd be huge delays writing data sometimes.
While it flushes it all to hdd?
yeah i think so
there was really high cpu
i can't remember what process
The theory is that most workloads are very bursty rather than constant throughput, so there is idle time that the OS can use to flush data back to the hdds without causing any impact
yeah most are
Good morning arp
mercutio: nah, i don't think i've ever seen a single trade request. i don't play tf2 anymore, so i guess that might limit peoples' interest in trading
good morning phlux
apparently BGPmon has been acquired by OpenDNS. Does anyone care?
oh wow.  i care.
well a little :)
you probably found this already, but https://blog.opendns.com/2015/03/12/opendns-acquires-bgpmon/
nah wasn't up with the play :)
opendns screws up cdn's from here.
seemingly even more so than google
how can I tell if my graphics card supports a 4k display?
GTX 750 Ti
nvm http://www.geforce.com/hardware/technology/4k/supported-gpus
mnathani_: Also, http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-750-ti/specifications
careful about 4k, some cards have mini displayport rather than displayport
and lots of monitors want to use displayport for 60 hz, and it seeems to be hard to go from mini to full whilst maintaining 4k support
the jump from 1080p to 1440p is bigger than 1440p to 2160p for desktop use imo
i swapped monitors back to 4k for linux now as my windows graphics card seems to lose sync with the monitor regularly.
but from what i understand that it's a general "r9 290" issue.
r9 290 is that an amd card?
yes
does this look decent? http://www.newegg.ca/Product/Product.aspx?Item=N82E16824009659&cm_re=4k_monitor-_-24-009-659-_-Product
all the current cheap ones are TN
it's anti glare rather than matte
depends on your room, but i prefer matte screens normally
there's thick and thin matte
That's what she said!!
a cursory look seems to suggest it's about ok, but it's more expensive than mine was
http://www.newegg.com/Product/Product.aspx?Item=N82E16824160203
this is what i have
it seems to be rated 3/5
my requirements going in were cheap, decent stand, 60 hz, no ugly silver crap
but i was a bit indecisive if i cared about high input lag etc.
the two main issues i have with the monitor are high input lag, and the menus are god awful to use.
but there's also worse colours than my 1440p pls monitor
it doesn't really matter for text though
http://www.ebay.com/itm/New-QNIX-QX2710-LED-Evolution-27-WQHD-PLS-2560x1440-Computer-Monitor-Matte-/400821450762?pt=LH_DefaultDomain_0&hash=item5d52d1f80a
i have one of these as my 1440p
which overclocks to 96 hz fine.
with an aoc stand on it
had it before 4k was cheap though