|
Shinku ABOOKEN posted:if what bsd is saying is true then how does swap/suspend/hibernate work on os x? the fixed partition is needed for kdump not suspend/hibernate kdump is running after the kernel has already crashed. that would be a really bad time to touch filesystem data structures, so it's important to already have a fixed swap location before it runs
|
#
?
Oct 27, 2014 21:39
|
|
|
|
| # ? Jan 23, 2025 18:29 |
|
|
Notorious b.s.d. posted:the only time you should ever touch swap, ever, is to dump kernel memory to disk. either for debugging, or for suspend/resume. this requires a fixed swap partition of known size to be immediately available w/out any code executing to create it or if you cj'd OpenVZ stacks during the time there was that kernel bug that wouldn't free swap, and you'd have to do the swapoff/swapon dance while people complained they can't SSH to the box. THANKS FOR YOUR OPTIMIZED KERNEL DICK HOLES!
|
|
#
?
Oct 27, 2014 21:42
|
|
|
Winkle-Daddy posted:OPTIMIZED KERNEL DICK HOLES mods? 
|
|
#
?
Oct 27, 2014 21:43
|
|
|
Winkle-Daddy posted:or if you cj'd OpenVZ stacks during the time there was that kernel bug that wouldn't free swap, and you'd have to do the swapoff/swapon dance while people complained they can't SSH to the box. THANKS FOR YOUR OPTIMIZED KERNEL DICK HOLES! virtuozzo is an abomination if i ever have to cat /proc/user_beancounters again it will be too soon
|
|
#
?
Oct 27, 2014 21:44
|
|
|
Shinku ABOOKEN posted:if what bsd is saying is true then how does swap/suspend/hibernate work on os x? uses files
|
|
#
?
Oct 27, 2014 21:53
|
|
|
sorry i meant to ask : can os x produce kernel dumps? is the regular os x's kernel debuggable?
|
|
#
?
Oct 27, 2014 22:07
|
|
|
Notorious b.s.d. posted:the only time you should ever touch swap, ever, is to dump kernel memory to disk. either for debugging, or for suspend/resume. this requires a fixed swap partition of known size to be immediately available w/out any code executing to create it as usual your reading comprehension is creatively stupid, bsd stymie also lol that u think the only purpose of swap is to dump kernel memory. even in environments where all the users are theoretically clueful, the real world is imperfect. having swap as a backstop prevents terrible problems like losing a week's worth of simulation (aka a week's worth of deadline) to the oom killer when some rear end in a top hat doesn't check whether it's safe to start up another sim on the same compute server
|
|
#
?
Oct 27, 2014 22:08
|
|
|
Shinku ABOOKEN posted:sorry i meant to ask : can os x produce kernel dumps? is the regular os x's kernel debuggable? yes, it can dump over tcp/ip. it also supports two-machine debugging where gdb on one system uses firewire remote dma to debug the live kernel on the other
|
|
#
?
Oct 27, 2014 22:12
|
|
|
most of the current programming talent goes to Apple to fix errors older than themselves.
|
|
#
?
Oct 27, 2014 22:16
|
|
|
BobHoward posted:even in environments where all the users are theoretically clueful, the real world is imperfect. having swap as a backstop prevents terrible problems like losing a week's worth of simulation (aka a week's worth of deadline) to the oom killer when some rear end in a top hat doesn't check whether it's safe to start up another sim on the same compute server lol you use swap as a backstop because you have no batch processing system. great job, guys
|
|
#
?
Oct 27, 2014 22:17
|
|
|
i think i have figured it out: bobhoward works in the late 1990s
how does the internet reach from us to your time, bob
|
|
#
?
Oct 27, 2014 22:18
|
|
|
Notorious b.s.d. posted:i think i have figured it out: bobhoward works in the late 1990s sounds more like you don't live in the real world but OK
|
|
#
?
Oct 27, 2014 22:20
|
|
|
Captain Foo posted:sounds more like you don't live in the real world but OK none of us do. we live in the 'pos
|
|
#
?
Oct 27, 2014 22:24
|
|
|
i live in a world where operations guys get paged immediately if a server hits swap because that is an error condition
|
|
#
?
Oct 27, 2014 22:25
|
|
|
googling for linux dynamic swap shows something like 3 projects that do this, apparently.
|
|
#
?
Oct 27, 2014 22:29
|
|
|
2014, year of Linux on the DESKTOP, where noone in their right mind gives / should give a poo poo about swap or ram utilization.BobHoward posted:and it's actually dumb everywhere, not just desktop Notorious b.s.d. posted:i live in a world where operations guys get paged immediately if a server hits swap sshhhhhhhhhhhhh
|
|
#
?
Oct 27, 2014 22:34
|
|
|
Notorious b.s.d. posted:i live in a world where operations guys get paged immediately if a server hits swap Is it cold in your ivory tower BSD stymie ?
|
|
#
?
Oct 27, 2014 23:06
|
|
|
jre posted:Is it cold in your ivory tower BSD stymie ? if it gets cold i call facilities and a nice man from Johnson Controls adjusts the HVAC to suit my needs
|
|
#
?
Oct 27, 2014 23:34
|
|
|
The solution is to buy more ram haha
|
|
#
?
Oct 27, 2014 23:48
|
|
|
|
|
#
?
Oct 27, 2014 23:48
|
|
|
I hope you all progress far enough in your funny computing careers that you reach the point of nirvana where you no longer give a poo poo about arguing over whether your devs do or don't get paged when a server hits swap holy Christ you are all spergs
|
|
#
?
Oct 27, 2014 23:55
|
|
|
 if you have one hundred and twenty eight gigabytes of random access memory in your personal computer architecture based server and you hit swap then you have some serious problems also loving lol you're saying a SIGBUS or a SIGSEGV would wipe out an entire week's worth of computing work? like seriously? you have absolutely no way to checkpoint this poo poo to disk? "doctor it hurts when i hammer myself repeatedly in the balls"
|
|
#
?
Oct 28, 2014 00:26
|
|
|
anyway as for my laptop uuh i don't want to flashburn my ssd when swapping and even with this 2009-era c2d piece of crap with 4gb of ram my computing needs seem to be fully suited without any sort of swap. i can see how cad or video editing or poo poo like that might actually need some decent amounts of ram tho but i mean at that point you probably would want IT to get your needs professionally suited for you and yeah swap files work fine anyway and have done since like ever not if you use btrfs mind you but btrfs is bad anyway imo
|
|
#
?
Oct 28, 2014 00:28
|
|
|
Mr Dog posted:anyway as for my laptop uuh i don't want to flashburn my ssd when swapping and even with this 2009-era c2d piece of crap with 4gb of ram my computing needs seem to be fully suited without any sort of swap. i can see how cad or video editing or poo poo like that might actually need some decent amounts of ram tho but i mean at that point you probably would want IT to get your needs professionally suited for you years ago i used my eee pc 701 with no swap as i had installed 4GB of ram which was coincidentally the size of the SSD in the laptop and it suited my needs fine
|
|
#
?
Oct 28, 2014 00:51
|
|
|
Captain Foo posted:sounds more like you don't live in the real world but OK
|
|
#
?
Oct 28, 2014 00:51
|
|
|
PIGEOTO posted:I hope you all progress far enough in your funny computing careers that you reach the point of nirvana where you no longer give a poo poo about arguing over whether your devs do or don't get paged when a server hits swap holy Christ you are all spergs spergs? in MY yospos!?
|
|
#
?
Oct 28, 2014 01:23
|
|
|
Shinku ABOOKEN posted:sorry i meant to ask : can os x produce kernel dumps? is the regular os x's kernel debuggable? i thought osx used the filesystem, an idiot choice, but i was wrong osx actually has a semi-sane design. you configure a second dump server on the network and it writes over the network. a crude solution but it should work ok. (what happens when i need to work on a network driver crash? linux ameliorates this stuff by having kdump execute a second kernel with un-crashed drivers)
|
|
#
?
Oct 28, 2014 01:25
|
|
|
Mr Dog posted:
lol if u think eda cad tools are always great about checkpointing etc. welcome to the world of being an end user who(se employer) pays huge amounts of money for the privilege of using software that is poo poo (okay this is everyone's world but it is especially true in my field) lol that nbsd thinks small companies in shaky condition (i'm no longer employed there due to corporate existential crisis  ) even have an on-call server ops teamlol that every time i post people keep ignoring the point which is that the most correct way to solve this problem (and no matter how much u think its not a real problem its still something os designers must worry about) is to use a simple technology from the stone age of computing u wouldn't accept static allocation of ram in a general purpose operation system why accept static allocation of disk for swap
|
|
#
?
Oct 28, 2014 01:45
|
|
|
BobHoward posted:why accept static allocation of disk for swap patches welcome etc
|
|
#
?
Oct 28, 2014 01:49
|
|
|
BobHoward posted:u wouldn't accept static allocation of ram lol someone never used basically any open source hypervisor
|
|
#
?
Oct 28, 2014 01:50
|
|
|
Notorious b.s.d. posted:i live in a world where operations guys get paged immediately if a server hits swap When I was an RHEL sysadmin, I could never ever figure out why exactly 128mb of swap started being used within 24 hours of a reboot. There was always at least 4GB of ram free. Considering every kernel errata for RHEL was novella length I figured it would be one of life's great mysteries. I envied the windows admin team. Reboot, Reboot, Reboot, Reboot. Fourth reboot didn't work? Call Microsoft.
|
|
#
?
Oct 28, 2014 02:16
|
|
|
pram posted:open source hypervisor n-n-nnn-n-n-no dont
|
|
#
?
Oct 28, 2014 02:16
|
|
|
vm.swappiness you moron http://en.wikipedia.org/wiki/Swappiness
|
|
#
?
Oct 28, 2014 02:17
|
|
|
keyvin posted:When I was an RHEL sysadmin, I could never ever figure out why exactly 128mb of swap started being used within 24 hours of a reboot. There was always at least 4GB of ram free. Considering every kernel errata for RHEL was novella length I figured it would be one of life's great mysteries. pram posted:vm.swappiness you moron incidentally this is why you monitor swapin/swapout rates and not swap usage directly there are innocuous ways to misconfigure a server that will cause small amounts of swap to be used, but it doesn't matter because rates will be 0: long-lived files in a tmpfs, vm.swappiness set too high and a long-lived app wires memory it doesn't access often etc
|
|
#
?
Oct 28, 2014 03:02
|
|
|
Notorious b.s.d. posted:i live in a world where operations guys get paged immediately if a server hits swap paged... swap... I GET IT
|
|
#
?
Oct 28, 2014 03:20
|
|
|
BobHoward posted:lol if u think eda cad tools are always great about checkpointing etc. welcome to the world of being an end user who(se employer) pays huge amounts of money for the privilege of using software that is poo poo i meant the guy who was doing week-long compute cluster simulations with no way to recover from a SIGSEGV or whatever http://en.wikipedia.org/wiki/CRIU this may be of interest to that person btw (still can't believe this insane poo poo exists)
|
|
#
?
Oct 28, 2014 03:50
|
|
|
Mr Dog posted:"doctor it hurts when i hammer myself repeatedly in the balls"
|
|
#
?
Oct 28, 2014 06:10
|
|
|
I solved my performance issues by moving my swap file to a RAM disk
|
|
#
?
Oct 28, 2014 09:47
|
|
|
More ram, haha
|
|
#
?
Oct 28, 2014 10:02
|
|
|
|
| # ? Jan 23, 2025 18:29 |
|
|
Soricidus posted:I solved my performance issues by moving my swap file to a RAM disk 
|
|
#
?
Oct 28, 2014 10:13
|
|