| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Comments:
<0> http://www.m90.org/index.php?id=11566 <-- this takeoff is..interesting
<1> Linus didn't design the scheduler
<0> speaking of which, arent you fairly free to choose what scheduler you want?
<2> blueroo: no, but he was the person who in the 2.4.x crap decided each thread should be a process
<2> "ohhh, let's give a thread a process id... yeah, so that people can kill -9 a thread".... BRILLIANT!
<1> er
<0> schitzo, meaning this is why when I do ps on a box running java it'll show several for actually just one process
<2> fredk_: bingo
<1> schitzo
<1> it has to be done that way
<1> unless you want to rewrite the entire scheduler *again*
<2> blueroo: it does? funny.. NT, Solaris and BSD don't do it that way
<2> it's a **** MODEL!
<2> someone was smoking crack
<2> none, but I write alot of thread applications and they **** on the 2.4.x stuff
<2> s/thread/threaded/
<2> 2.6.x is better. much better than 2.4.x
<2> we still have to duct-tape stuff to get it to work properly on 2.6.x, but not as bad as on the 2.4.x stuff
<1> imagine that. they rewrote the scheduler in 2.6 to improve flexibility and with threads in mind, and it improved
<2> yeah, lookie there
<3> Okay. Need some unusual yet highly cool things to do in San Francisco tomorrow.
<2> Evii: escape from alcatraz.
<0> Evii, drop by M5 industries?
<2> Evii: get a tattoo
<3> Bearing in mind this is with $hot_chick :)
<2> Evii: toss fish down at the wharf
<0> Evii, you telling me M5 Industries aint her cup of tea? :)
<2> Evii: go beat up hippies in berkeley
<3> schitzo: Do they eat human bones like pigs do?
<3> schitzo: The hippies, I mean.
<1> schitzo, you do realize that thread processes in linux are not the same as a regular process?
<3> fredk_: That could perhaps be what I was hinting at :)
<0> Evii, drive down to the desert and blow up ****?
<2> blueroo: yes, but I still should not in no way no how be able to kill -9 a thread
<3> fredk_: That's more like it.
<2> blueroo: but that has been fixed with the 2.6.x stuff
<2> Evii: jump off the golden gate bridge.
<1> if that's your only complaint, your argument is pretty weak
<3> schitzo: "Awesome" as we say here in California :P
<1> if anything, Linux threads are more flexible than Solaris threads
<2> blueroo: put down the crack pipe
<1> apparently you don't know that Linux allows you to choose which resources are shared when you clone a new thread
<1> or that viewing a thread as a task like any other task allows Linux to better balance it across CPUs
<2> then how come Solaris or BSD or NT doesn't need to do that?
<2> actually NT has the best threading out of anything
<2> as much as it pains me to say that
<1> they try to do all the footwork in a black box that you can't touch
<2> which is why it works so much better
<1> you're confusing "better" with "faster"
<1> NT is much faster, yes
<1> but there are trade-offs
<1> there are always trade-offs
<0> w00
<1> besides, can you really compare a system that has a 100% thread based scheduler to a system that shares threading and traditional processes?
<4> * Checking root filesystem ...
<4> fsck: fsck.xfs: not found
<4> fsck: Error 2 while executing fsck.xfs for /dev/sda3
<4> why?
<2> I know that 3000+ threads on NT or Solaris work much better than on a 2.4.x kernel machine that pretty much upchucks and dies
<2> fredk_: Vegas rawks
<0> schitzo, I know :)
<1> why are you still hung up on 2.4?
<1> 2.4 is old old news. get with the 2.6 program
<2> blueroo: because that is the one that ****s. 2.6.x like I said handles it much better
<2> blueroo: and my complaint about all of that was with teh 2.4.x kernel if you were paying any attention at the start of it
<1> ok, and Windows 3.11 for Workstations ****ed balls too
<1> you don't see anyone complaining about that
<2> now I just watch load avgs. climb for no reason what so ever
<2> yeah WFW did ****
<0> Like win 3.0 didnt
<1> load averages don't climb without a reason, btw
<1> there's always a reason
<2> heh, the operations folks don't seem to agree.
<2> and watching various machiens for a couple of months don't agree either
<2> the best part is I can shut down all applications on the box, remove all NFS mounts
<2> and the load will still sit there at 27
<2> it's wicked awesome
<1> computers are not independant thinking machines
<1> if something happens, it's for a reason
<2> but as fredk_ suggested. I don't have other hardware to test it on to see if it's something hardware related
<2> even though all monitoring options show the box to be doing nothing.
<1> what does vmstat say?
<2> 1 0 8 1417544 41168 2413696 0 0 0 0 1 0 4 1 95 0
<2> that's a load of 12.5 atm
<2> 4:44pm up 104 days 10:51, 1 user, load average: 12.50, 12.58, 12.64
<1> 95% system?
<2> to be exact
<2> 95 id
<2> 1 sy 4 us
<1> need a header then. your vmstat is slightly different
<2> r b swpd free buff cache si so bi bo in cs us sy id wa
<2> 1 0 8 1417544 41168 2413696 0 0 0 0 1 0 4 1 95 0
<0> schitzo, I ***ume this is some weird config or could I just boot a box and you install what you need remotely and let sit for three days?
<1> your tools are borked
<2> fredk_: it would need to take load and actually do shiut
<1> you're not getting any readings for i/o or context switching
<2> blueroo: hold on. just did a vmstat 5 5
<2> http://pastebin.ca/81750
<1> what does this machine do?
<2> webserver
<2> SunONE 6.1sp3, with a couple of java based web applications
<1> how much physical memory, how much swap?
<2> 4Gb memory 2 Gb swap
<2> Mem: 4146776k total, 2729556k used, 1417220k free, 41204k buffers
<2> Swap: 2104504k total, 8k used, 2104496k free, 2427024k cached
<2> If I shutdown the webserver, the load does not drop
<1> I'm curious about those 8 swapped processes
<2> the 8k of used swap?
<1> something isn't adding up
<2> well, the only appliction other than OS stuff that is running on the box is postfix and the webserver
<2> and I have no clue why postfix is there.
<2> otherwise all this machine does is HTTP traffic
<1> does your app send mail?
<2> nope
<1> take that up with your system design guys then :)
<2> HTTP GET to jsp. JSP uses Beans that make HTTP requests to third party for XML data. DAta is parsed and JSP creates HTML to send back to the user
<1> single proc machine?
<2> that's what it does
<2> no. two CPU with hyperthreading turned on
<2> Intel(R) Xeon(TM) CPU 3.20GHz
<1> that's an awful lot of context switching for such little cpu time
<2> lots of threads
Return to
#linux
or
Go to some related
logs:
#windows
what is going on with aaameds
123000220
#windows
lacheti
wrt54g 550 permission denied
#chatzone
#php
#php
deob java
|
|