High memory kernel support
I have a server that started throwing the following error the other day: kernel: __alloc_pages: 0-order allocation failed (gfp=0x1d2/0) We run some pretty memory intensive apps, and from looking at various logs it appears the machine ran out of memory space and started shutting down services. I was able to stop and restart all of the server daemons and free up memory. Since then everything seems to be fine. In researching the cause, I find that my kernel does not have high memory support compiled into it (the default). Now this is all a little over my head, so I'm asking here, what do you think, will recompiling the kernel to include high memory support resolve this? The machine had been up for 168 days before this happened. It has 1GB of physical RAM, and 2GB of swap space. Slackware 10.0. Thanks, --charlie -- Charles Farinella Appropriate Solutions, Inc. (www.AppropriateSolutions.com) [EMAIL PROTECTED] 603.924.6079 ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, 2005-03-17 at 10:47 -0500, Charles Farinella wrote: > I have a server that started throwing the following error the other > day: > > kernel: __alloc_pages: 0-order allocation failed (gfp=0x1d2/0) > > We run some pretty memory intensive apps, and from looking at various > logs it appears the machine ran out of memory space and started shutting > down services. I was able to stop and restart all of the server daemons > and free up memory. Since then everything seems to be fine. > > In researching the cause, I find that my kernel does not have high > memory support compiled into it (the default). Now this is all a little > over my head, so I'm asking here, what do you think, will recompiling > the kernel to include high memory support resolve this? The machine had > been up for 168 days before this happened. Compiling in high memory support will only help you if you plan to add more memory. If you have between 1GB and 4GB of RAM, then you want to set this to the 4GB option. > It has 1GB of physical RAM, and 2GB of swap space. Slackware 10.0. What kernel are you running? Kenny signature.asc Description: This is a digitally signed message part
Re: High memory kernel support
On Thu, 2005-03-17 at 11:03, Kenneth E. Lussier wrote: > > In researching the cause, I find that my kernel does not have high > > memory support compiled into it (the default). Now this is all a little > > over my head, so I'm asking here, what do you think, will recompiling > > the kernel to include high memory support resolve this? The machine had > > been up for 168 days before this happened. > > Compiling in high memory support will only help you if you plan to add > more memory. If you have between 1GB and 4GB of RAM, then you want to > set this to the 4GB option. Oh good, that will save me some work. :-) > > It has 1GB of physical RAM, and 2GB of swap space. Slackware 10.0. > > What kernel are you running? 2.4.26 -- Charles Farinella Appropriate Solutions, Inc. (www.AppropriateSolutions.com) [EMAIL PROTECTED] 603.924.6079 ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, 2005-03-17 at 11:40 -0500, Charles Farinella wrote: > On Thu, 2005-03-17 at 11:03, Kenneth E. Lussier wrote: > > Compiling in high memory support will only help you if you plan to add > > more memory. If you have between 1GB and 4GB of RAM, then you want to > > set this to the 4GB option. > > Oh good, that will save me some work. :-) > > > > It has 1GB of physical RAM, and 2GB of swap space. Slackware 10.0. > > > > What kernel are you running? > > 2.4.26 > Well, for many reasons, I would suggest upgrading the kernel to 2.6. And, of course, add another GB of memory :-) HTH, Kenny signature.asc Description: This is a digitally signed message part
Re: High memory kernel support
"Kenneth E. Lussier" <[EMAIL PROTECTED]> writes: >> > What kernel are you running? >> >> 2.4.26 >> > Well, for many reasons, I would suggest upgrading the kernel to 2.6. Mmmm, what reasons are there for upgrading to 2.6 at this point. I've thus far treated 2.6 as 'testing/unstable' kernel, and since many of it's nicer features (like SATA) are getting backported to the 2.4 series, I haven't had a reason to venture into 2.6 land yet. -- Seeya, Paul ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
Paul Lussier wrote: Mmmm, what reasons are there for upgrading to 2.6 at this point. I've thus far treated 2.6 as 'testing/unstable' kernel, and since many of it's nicer features (like SATA) are getting backported to the 2.4 series, I haven't had a reason to venture into 2.6 land yet. I'll bite: drivers do not a kernel make. Many of the subsystems (eg., memory management, kernel locking) have been substantially refined, and I doubt that all the enhancements in, say, SATA are being backported. In addition to that, IMHO, the 2.6 kernel is, BY FAR, the most stable .0 (and subsequent) release I've ever seen. Gone are the days of the 2.4.9 debacle, the 2.2.0 debacle, etc. I've been running with nary a glitch since 2.5.58 (?) or so, and have upgraded all my important servers to 2.6 back at 2.6.3. I doubt it's 100% BugFree(tm), but I do think that, kind of in opposition to your stance, unless there's a reason to stick with 2.4 (such as, "It just works," which is a fine reason), 2.6 is the way to go. $.02, etc., -Ken ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, Mar 17, 2005 at 05:27:25PM -0500, Ken D'Ambrosio wrote: > In addition to that, IMHO, the 2.6 kernel is, BY FAR, the most stable .0 > (and subsequent) release I've ever seen. Gone are the days of the 2.4.9 > debacle, the 2.2.0 debacle, etc. It just goes to show that, as with everything else, YMMV. My experience has been quite diferent. I've had innumerable crashes and oopses related to USB in particular... -- Derek D. Martinhttp://www.pizzashack.org/ GPG Key ID: 0xDFBEAD02 -=-=-=-=- This message is posted from an invalid address. Replying to it will result in undeliverable mail. Sorry for the inconvenience. Thank the spammers. pgpv7hwfyuF9A.pgp Description: PGP signature
Re: High memory kernel support
Derek Martin <[EMAIL PROTECTED]> writes: > On Thu, Mar 17, 2005 at 05:27:25PM -0500, Ken D'Ambrosio wrote: >> In addition to that, IMHO, the 2.6 kernel is, BY FAR, the most stable .0 >> (and subsequent) release I've ever seen. Gone are the days of the 2.4.9 >> debacle, the 2.2.0 debacle, etc. > > It just goes to show that, as with everything else, YMMV. My > experience has been quite diferent. I've had innumerable crashes and > oopses related to USB in particular... My mileage does vary. USB has been quite a bit more stable for me with 2.6 than with 2.4. --kevin -- GnuPG ID: B280F24E And the madness of the crowd alumni.unh.edu!kdc Is an epileptic fit -- Tom Waits ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, Mar 17, 2005 at 06:05:29PM -0500, Kevin D. Clark wrote: > > Derek Martin <[EMAIL PROTECTED]> writes: > > > On Thu, Mar 17, 2005 at 05:27:25PM -0500, Ken D'Ambrosio wrote: > >> In addition to that, IMHO, the 2.6 kernel is, BY FAR, the most stable .0 > >> (and subsequent) release I've ever seen. Gone are the days of the 2.4.9 > >> debacle, the 2.2.0 debacle, etc. > > > > It just goes to show that, as with everything else, YMMV. My > > experience has been quite diferent. I've had innumerable crashes and > > oopses related to USB in particular... > > My mileage does vary. USB has been quite a bit more stable for me > with 2.6 than with 2.4. For me, it depends upon exactly which 2.6 version I'm using... I'm currently sticking with Fedora's 2.6.8-1.521 kernel, because the 2.6.9-X and 2.6.10-1.9_FC2 updates gave me problems with all of my USB devices. Specifically, when I connected an/or disconnected them, the kernel would frequently (but not always) oops, and the system would eventually stop working after that (not surprisingly). If I tried to shut down at any time after the oops, the kernel would invariably get wedged before the shutdown was complete. I think the 2.6.10-1.12 update was better, but I discovered that my (win)modem driver won't work with the 2.6.10+ kernel, so I haven't really had opportunity to test it extensively. -- Derek D. Martinhttp://www.pizzashack.org/ GPG Key ID: 0xDFBEAD02 -=-=-=-=- This message is posted from an invalid address. Replying to it will result in undeliverable mail. Sorry for the inconvenience. Thank the spammers. pgpnTV1NIG1ul.pgp Description: PGP signature
Re: High memory kernel support
Ken D'Ambrosio <[EMAIL PROTECTED]> writes: > but I do think that, kind of in opposition to your stance, unless > there's a reason to stick with 2.4 (such as, "It just works," which > is a fine reason), 2.6 is the way to go. Well, our product is based currently based on the 2.4 series, so there's significant resistance to moving towards 2.6 because of that. Also, we try to run the same kernel not only on the product, but also everywhere else; desktops, servers, etc. That way we have the widest possible use-case of the kernel. If there's a bug, we're using the kernel in so many different ways, that hopefully we'll find it before it turns up at a customer site. I didn't mean to imply *I'm* hesitant to upgrade, just that I haven't found an overly compelling need to do so. Keeping in mind that to propose shifting would also require me to convince all the various "powers that be" we shouldshift. That's sometimes a lot more effort than it's worth ;) -- Seeya, Paul pgpW8IaZcgC9D.pgp Description: PGP signature
Re: High memory kernel support
On Thu, 2005-03-17 at 17:05 -0500, Paul Lussier wrote: ... > Mmmm, what reasons are there for upgrading to 2.6 at this point. I've > thus far treated 2.6 as 'testing/unstable' kernel, and since many of > it's nicer features (like SATA) are getting backported to the 2.4 > series, I haven't had a reason to venture into 2.6 land yet. I use 2.6 in a number of servers and workstations, and it's as stable as ever. I build them from the kernel.org releases. My problem is that some of the drivers I need aren't keeping up with the kernel releases, like the NVIDIA drivers, for instance. Still, if you are running mission critical, you have to make the best decision for your situation. There are issues with certain driver/kernel combinations, but usually there are workarounds. -Fred ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
> My problem is that some of the drivers I need aren't keeping up with the > kernel releases, like the NVIDIA drivers, for instance. The driver issue in 2.6 has impacted me, though in the other direction. 2.6 includes some new mouse drivers and I've had issues with those in oldish touchpad laptops (though more recent releases have been better). Regards, . Randy -- Bill Gates' wealth equals the combined wealth of the poorest 120 million Americans, or 45 percent of our population. "This is a failure of the political system to defend the people." -- Ralph Nader ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, 2005-03-17 at 23:18 -0500, Randy Edwards wrote: > > My problem is that some of the drivers I need aren't keeping up with the > > kernel releases, like the NVIDIA drivers, for instance. > >The driver issue in 2.6 has impacted me, though in the other direction. > 2.6 includes some new mouse drivers and I've had issues with those in oldish > touchpad laptops (though more recent releases have been better). I despise touchpads. I use a wireless mouse with my laptop. In any event, I used to have issues with touchpads under the 2.4 kernel. The 2.6 kernel has cleared up those issues for me. I would imagine a lot of backporting has taken place by now, anyway. Now, if we can just get NVIDIA to keep more up to date and do releases more often than once a quarter -- if that much. Back to the kernel -- what I'd like to do evenutally is build a 64-bit machine with perhaps 16Gbytes of ram, and make /tmp a ramdisk so I can store all sorts of potentially sensitive information there that I don't wish to keep on the harddrive, such as file caches and the like. Also, I want to eliminate the paging file altogether -- don't see a point for it when you have that much RAM. Are there any issues with having a kernel running without a paging file outside of giving you "extra" memory to swap out least-used pages? Also, are there any issues with getting 32-bit drivers to work in a 64- bit environment? Especially proprietary ones like NVIDIA? -- -Fred Your government delights in doing things "for" you. By the time it's too late, you'll realize it's being done *to* you. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: High memory kernel support
On Thu, 2005-03-17 at 17:05 -0500, Paul Lussier wrote: > "Kenneth E. Lussier" <[EMAIL PROTECTED]> writes: > > Mmmm, what reasons are there for upgrading to 2.6 at this point. I've > thus far treated 2.6 as 'testing/unstable' kernel, and since many of > it's nicer features (like SATA) are getting backported to the 2.4 > series, I haven't had a reason to venture into 2.6 land yet. In the driver sense, I originally moved to 2.6 because 2.4 was lacking drivers for GigE cards such as the Intel 10/100/1000 (which I believe has since been back ported). Also, the SCSI sub-system has been re-vamped to make it more secure and robust (and break cd writing in some cases). There is also the cryptography support. Gone are the days of having to patch the kernel for IPSec. As well as the usual refinements and fixes that one would expect to see in the newer kernels. However, there have been some issues. I had some problems with 2.6.8 and 2.6.8.1. But those issues were fixed two days later when 2.6.9 came out. FYI, Kenny signature.asc Description: This is a digitally signed message part
Re: High memory kernel support
On Fri, Mar 18, 2005 at 08:23:47AM -0500, Kenneth E. Lussier wrote: > There is also the cryptography support. Gone are the > days of having to patch the kernel for IPSec. Interesting indeed. I recently got wireless working on both my laptops, though I don't currently have any enryption going on the wifi network. I believe the Linux driver for the Intel 2100 doesn't support WPA, meaning the best I could do anyway is WEP, which isn't very good. I was thinking about using IPSec to solve that... Are you using IPSec now? I heard that FreeS/Wan forked... What code base are you using? Pointers would be useful. =8^) -- Derek D. Martinhttp://www.pizzashack.org/ GPG Key ID: 0xDFBEAD02 -=-=-=-=- This message is posted from an invalid address. Replying to it will result in undeliverable mail. Sorry for the inconvenience. Thank the spammers. pgpow6fMj7iKF.pgp Description: PGP signature
Re: High memory kernel support
On Fri, 2005-03-18 at 10:15 -0500, Derek Martin wrote: > On Fri, Mar 18, 2005 at 08:23:47AM -0500, Kenneth E. Lussier wrote: > > There is also the cryptography support. Gone are the > > days of having to patch the kernel for IPSec. > > Interesting indeed. I recently got wireless working on both my > laptops, though I don't currently have any enryption going on the wifi > network. I believe the Linux driver for the Intel 2100 doesn't > support WPA, meaning the best I could do anyway is WEP, which isn't > very good. I was thinking about using IPSec to solve that... WPA, 802.1x authentication, etc. requires a few things in addition to the driver. Check out http://hostap.epitest.fi/wpa_supplicant/ for more info on that. > Are you using IPSec now? I heard that FreeS/Wan forked... What code > base are you using? Pointers would be useful. =8^) I haven't done much with IPSec recently (in the last 3 months). However, when FreeS/WAN shut down due to political reasons, I started looking into OpenS/WAN http://www.openswan.org/. I was impressed by the rapid development on the project. It has developed much faster then FS did because it isn't encumbered by the political and ideological issues that FS had. There is also a better effort for client interop. HTH, Kenny -- Kenneth E. Lussier Sr. Systems Administrator Sentito Networks signature.asc Description: This is a digitally signed message part
Re: High memory kernel support
Derek Martin <[EMAIL PROTECTED]> writes: > Are you using IPSec now? I heard that FreeS/Wan forked... What code > base are you using? Pointers would be useful. =8^) OpenBSD.org :) It's the most secure, most stable OS out there right now, and IPSec is built-in to the default kernel. I'm in the process of building a VPN concentrator right now, and it's turning out to be pretty simple with OpenBSD. Far more so than if I had used Linux. -- Seeya, Paul pgpbnfXhxTrrx.pgp Description: PGP signature
Re: High memory kernel support
On Fri, Mar 18, 2005 at 09:15:16PM -0500, Paul Lussier wrote: > Derek Martin <[EMAIL PROTECTED]> writes: > > > Are you using IPSec now? I heard that FreeS/Wan forked... What code > > base are you using? Pointers would be useful. =8^) > > OpenBSD.org :) > > It's the most secure, most stable OS out there right now, and IPSec is > built-in to the default kernel. I'm in the process of building a VPN > concentrator right now, and it's turning out to be pretty simple with > OpenBSD. Far more so than if I had used Linux. All of that may be true, but at the moment I've no desire to run OpenBSD... Even if I did, given my current situation, it's a practical impossibility. At some point I want to get cozy with the BSDs, but today is not the day... -- Derek D. Martinhttp://www.pizzashack.org/ GPG Key ID: 0xDFBEAD02 -=-=-=-=- This message is posted from an invalid address. Replying to it will result in undeliverable mail. Sorry for the inconvenience. Thank the spammers. pgpCNnGYjH843.pgp Description: PGP signature