Journal tomhudson's Journal: Bye-bye OpenSUSE 12.1, Hello Fedora 16 35
Well, since a second upgrade attempt left opensuse even weirder (dialogs with half the controls not working, firefox going from crashing every second load to every load, etc) it was time to nuke it again, but this time replace it with something different.
The question was, what?
It turns out that FreeBSD does not like my video setup (which is too bad, because I had 6 consoles open, and compiling a different part of the ports tree in each one, and there was no indication that it was under any sort of load, even though the load average was ~6).
Linux Mint? Tempting, very tempting
Good old slakware? I downloaded the DVD (using knoppix, since the os was hosed), then went looking for updates
So, what the heck - go grab Fedora 16 and install it
I finally figured out the problem - for some reason it doesn't see my usb keyboard (plugged into my screens' usb hub) and it's waiting for a keyboard to appear
The funny part - I've always found gnome to be kind of ugly, but the old 2.whatver gnome, the way they fixed it up is nice. I could get used to it
the evil part
SElinux. I removed it, and the machine is MUCH faster. so when they say it "only used ~7%" I'm not buying it.
In other good news, my colour laser FINALLY WORKS!!!!!! It was recognized, but no drivers - and this time the Samsung drivers installed with no hassles, so the only thing that still doesn't work is the scanner. I can always scan to usb (or maybe just make a patch cord and plug it right into the computer that way????)
One last speed-up
If you looked at my previous post, I tried to load it down, opening eclipse, openoffice, the gimp, playing mp3s in amarok, firefox and opera both open, web server, ftp server, mail and news servers running in the background
Just goes to show that the real bottlenecks for everyday use are mostly self-inflicted "best practices" that aren't so great any more. If you want to try your machine w/o swap, but have the option of restoring it, just fdisk the drive and change the partition type from swap to anything else - no need to format it, since you won't be mounting it. If it runs okay, then you might want to reformat it and use it as a separate
excuse me, Mr Miyagi (Score:1)
Couldn't you test the speed up this way?
swapon, swapoff
Re: (Score:2)
Re: (Score:1)
I try and avoid poking things I shouldn't touch, I only discovered swap/off after I broke something playing with something else.
Haven't had to look in mtab in a long time and have avoided it successfully.
So... (Score:2)
You've evidently got a hangup about upgrades.
When are you going to learn that upgrades are a complete waste of time?
Backup ~/, wipe, reinstall is the only sane way to do it.
Oh, you just did. Note, however, that the changing your distro part is optional.
Med vänliga hälsningar,
Z
Re: (Score:2)
OpenSUSE 12.1 works for me! (Score:2)
I think you're doing something wrong, or your hardware is seriously fucked.
Reason: I am running OpenSUSE on a Mac Pro at the office, and on a Dell Precision M6400 Mobile Workstation. Both work perfectly and are stable out of the box. Everything on the notebook works (well, I haven't tested the webcam since installing OpenSUSE 12.1, but it worked in previous versions): bluetooth, audio, wifi, the automatic brightness adjuster for the screen, keyboard backlighting, eSATA, USB, integrated cardreader, and so on
Re: (Score:1)
Re: (Score:2)
I ran memtest86+ yesterday "jus
Re: (Score:2)
Wait a sec - I seem to recall your complaints starting on the day of release - for the past few releases. One thing I learned in late 90s was never, ever install a network-updatable Linux distro on the day of release. The few times I've done so it's bitten me with various glitches. I always give it at least a couple of weeks so that any packages that the release engineer or maintainer might have slipped in without their having been tested by the sponsor's QA team or by the public during beta are found and
Re: (Score:2)
Both an in-place update and a fresh install failed, leaving me with having to re-install a fresh copy of 11.4. That copy ha
OOM? (Score:2)
So, what happens to your machine when you run out of physical memory to run all that stuff? I can burn through 4 gigs on my dev workstation when running a copy of the 2 apps I'm working on and running integration tests against them in Eclipse. Without swap to page out some of the memory the whole machine would just go "out of memory, I give up" and die a horrific death on me. I can't see running with no swap partition mounted when you need to fully utilize all the resources on the machine.
Re: (Score:1)
Yes, Rusty is halfway right here: I will tell you what happens if you run out of memory on a Linux system. A process is going to be shot down to free up memory. I'm not entirely sure if it's random (that should be googleable), but it makes for very interesting behaviour. Especially on servers. We had a VPS acting very strangely, each Sunday at 6h15 or so. Turned out, the cron jobs started (weekly), ate up too much memory. Sometimes apache got shot down, sometimes ntpd, sometimes something else entirel
Re: (Score:2)
The disks are fine, the ram is fine, and the problems arose on both the laptop and the desktop at the same time, due to my habit of installing all upgrades pretty much every day (note - in cases of conflicts, I would not force an install or upgrade, so in theory, nothing should have broken. Theory and practice, however, are two different things).
smartctl says the disks are fine. Ditto for memtest86+ and the ram. Also, I'm well aware of the difference between virtual memory address remapping in ram and
Re: (Score:2)
I'm a bit puzzled, in that I almost never see swap partition space actually used. A year or so ago, I noticed my old computer with only 1 GB of RAM had swap disabled, due to an issue with distribution upgrade scripts, so I'd had swap disabled for something like six months without noticing, even though I was doing things like running VMs on that box. At work, in our NOC, if servers start using more than a few percent of their swap partition space, we get alerts; but that doesn't happen often. So my sense is
Re: (Score:1)
Well, that stems from the idea that the swap partitions aren't used. They are. You think they are empty because that's what they are reported to be. In reality, data is written to them. Data that is in memory at this moment. At the moment memory needs to be freed up, the data in memory is simply marked free and the swap is already ready. This saves time when real memory is needed. That's why, if you can, you should always have at least as much swap space as there is real memory.
Most people think th
Re: (Score:2)
If swap partitions are used for paging, than how could systems without swap partitions even boot? Yet from tomhudson's account, she gets better performance without swap at all.
And from my experience, due to a conflict between a distribution upgrade script and the way Ubuntu treated Windows partitions in fstab during installation, I and presumably many other Ubuntu users had swap disabled for months, and almost no one noticed. On a low memory system, I had swap disabled, and I couldn't find any difference in
Re: (Score:2)
Agreed - swap files are simply not used for caching disk reads. That would be crazy - read from one area of the disk, then write a copy back to another area rather than just re-read the original sector(s) again? Writes are expensive! Even the most optimal writes need to match these condtions (1) you bypass the filesystem and go directly to a sector and (2) you can write exactly n*sector-size bytes and (3) your data starts and ends on sector boundaries.
Nothing beats real physical ram. Best is cpu regis
Re: (Score:2)
You forgot to mention SSD between cache on hard disks and rusty bits on spinning platters.
Re: (Score:2)
What will beat both is real ramdrives - drives that are just lots of ram and a battery to keep it alive - but they're a lot more expensive, and you still need the rust-boxes to serve as stable storage.
BTW - there's another meme that we're going to have to kill off as well - the idea that you should u
Re: (Score:2)
However, stuff that has been changed in memory (dirty) cannot be re-read from the original sectors of the drive, so you have to save it to the swap partition before overwriting that memory page with new bits.
There's a pathological problem with the linux kernel swapping out stuff when memory is still available, "j
Re: (Score:2)
1. Get more ram. There's no substitute for real physical ram, and the speed-up will more than pay for the cost - even if it means buying a new machine.
2. Re-write your tests so they use less ram. They'll probably run faster in series if you're not always swapping out anyway than if you run them concurrently and swap out all the time.
3. Make your tests stand-alone, instead of running through eclipse. This way, you can "pawn them off" to another machine and find pot
Re: (Score:1)
No, it's not... Go back at reading the concept of Virtual Memory and how it fits in the whole concept. Swap is nothing more than another level of cache, just a very slow one. Please, instead of keeping these prejudices, go and read it up. Paging is an important part of virtual memory implementation in most contemporary general-purpose operating systems [wikipedia.org]. You're doing nobody a favour by keeping your prejudices. You'll also see
Re: (Score:2)
Even the cheaper option - $85 for 12 gigs or ram, is overkill for anyone I know. Why bother with swap under those conditions?
Swap made sense when ram was $100 for 64k. It made sense when ram was $100 per meg. It made a bit less sense when ram was $100 per gig. It doesn't make any sense whatsoever when ram is $6 to $12 a
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
Fair enough. But you're right :-)
Re: (Score:1)
Re: (Score:2)
On a more serious context, this is the second time I've set up a distro w/o swap, and there are no issues, and obviously no speed lost in swapping. For cases where you just don't have enough physical ram, then obviously, use swap - it's better than doing nothing.
But the research on microkernels revealed that the biggest, #1 reason they were so slow was because they devoted entirely too much memory to swap, and in a microkernel, where you have to do several more context swi
Re: (Score:2)
You never know until you actually try it in real life, because reality bites :-)
Re: (Score:2)
Oh, I have done. And I mostly agree with you. RAM is cheap, and it makes sense to take advantage of that. But no matter how much RAM you have, you have to give some thought to what you want to happen when you run out. Because you will. If letting the OOM killer nuke the process it thinks is most problematic is acceptable, then fine. But there are plenty of situations where that's not OK a
Re: (Score:2)
Why get performance out of what we have? Spend more and replace it instead!
Sorry Barbra, but I just got this Macbook Pro. It has 8 gig of ram and is max'ed out. The application I am testing requires a 4+ gig heap under light use. The controller web-app requires 1.5 gigs of heap and the VM to run an Oracle DB server needs another 2 gigs. That leaves 512 megs for the OS, my debugging tools and the test suite to execute in. Without swap something would fail or I would not be able to run the VM which allow
Re: (Score:2)
Even a few years ago, it was possible to figure out that ram prices would drop to the point where 32, 64, or even (for developers working on "really interesting stuff":) 256 gig would be feasible, and that laptops should be able to take at least 16, and preferably 32 gigs .. stinkpads have been taking 16 gigs for years, and you
SELinux (Score:2)
Possibly the worst decision you'll ever make when it comes to Unix sysadmin. SELinux is an absolute essential on any box I own. I can't see why anyone would trade security for a minor[1] performance gain.
[1] From my experience, 7% is a massive overstatement, and I'd say it's closer to 1% or 2%. But even if you're right and it's much more than that. Say 25% (which it isn't). I'd still say it's worth it...
Re: (Score:2)
The 7% is the SELinux projects' figure, not mine ... and I believe it, because it was a real PITA.
I'll trade a minuscule reduction in security for a much increased usability any day ... but then again, I do have backups, so I'm not walking a tightrope without a safety net.
It's not like this is a server or anything ...
Re: (Score:2)