Re: corruption during power loss
Michael Soulier wrote: My biggest complaint about RedHat/Mandrake while I was using them was the fact that if I lost power, the disk caching would cause the filesystem to be corrupted, often seriously so. I'd cringe when I booted up again, because inevitably, I'd be prompted to login as root and run fsck myself. Often I'd have to reinstall a bunch of packages, and it didn't help that doing an rpm -Va to verify everything inevitably returned false negatives on package integrity due to broken rpms. Then I installed Debian. I've had about 5 losses of power since, and each time, the system has come up without a scratch. Minor inode problems easily fixed by fsck without manual control necessary. I noticed the entries in /etc/inittab for powerloss, but the script it's pointing to for me is not installed, so it's not that, although I'd like to know what this /etc/init.d/powerfail script is. So I ask, what aspect of Debian makes it superior in this respect? What's causing the wonderful lack of corruption during power failures? i'd highly reccomend getting a decent UPS. they are quite cheap, and they will not only prevent curroption during outages but it will make the system more stable(its usually difficult/impossible to notice spikes/brownouts by the time the computer shows signs it may be too late..) the life of the hardware will be improved, and a lot of headaches will go away. you don't need a monitor on a UPS(unless your me) but a computer without a UPS is like well..asking for trouble. my systems ... my 2 main systems run off of cyber power 2200s (1100VA) they provide about an hour worth of backup time(at the time the UPSs were being phased out and i got them each for $99 w/free overnight delivery, shoulda gotten 3-4 more) i have a P5-200 on a cyberpower 320VA(got it for $39) got my tv/vcr/cable on another 320VA cyberpower and got my cordless phone on ANOTHER 320VA ..i have a APC back ups 400 but the battery is dead its useless till i get around to replacing it :) i can't stress enough a UPS to anyone that is even slightly serious about computers no matter what OS or filesystem they run, the risk to damaging hardware(IMO) is greater then the risk to damaging software. you are in canada so you may not be affected by the serious power problems the U.S. is currently experiencing(wide spread black/brown outs). because of the deregulation of the power industry here there hasn't been a new power plant built in more then 5 years, infact many have closed down(i can think of at least 1 nuclear power plant that closed where i used to live). the situation is going to get worse before it gets better. i didn't realize how many power problems i actually have until i had a UPS that beeped at every one. UPSs are not just for servers, they should be just as standard as a keyboard or a mouse to a PC. kinda sad that people don't take it seriously enough. and btw UPSs are really more useful for cleaning up dirty power rather then protecting against full blown blackouts. ive seen many instances where APC smart UPSs registered severe voltage drop offs(going from 121V to as low as 50-60VA), may not be enough to kill the lights, but probably enough to kill your PC(s). my 0.02 nate -- ::: ICQ: 75132336 http://www.aphroland.org/ http://www.linuxpowered.net/ [EMAIL PROTECTED]
Re: corruption during power loss
On Thu, Sep 14, 2000 at 11:51:01PM -0400, Michael Soulier wrote: On Fri, 15 Sep 2000, Pollywog wrote: About half the time that I experience power failures, I need to run fsck on my Debian system. So, have you had to reinstall packages? With RedHat/Mandrake, I would immediately do an rpm -Va to verify everything, but I quickly found out that several packages would complain about being broken even after a fresh install, so the information was useless. I'm still getting deeper into Debian's packaging tools. I tried running sync on a cronjob every hour, but it didn't help much. I never understood why Linux seemed so vulnerable to this. I've had the system crippled for hours with corrupt files while I reinstall packages. Not since installing Debian, but considering what you've said here, maybe I've been lucky. Try mounting your drive with the options noatime,sync and then install noflushd which will cause a sync/powerdown of drive when there is no activity. HTH Mike -- Unsubscribe? mail -s unsubscribe [EMAIL PROTECTED] /dev/null -- Juli-Manel Merino Vidal Email: [EMAIL PROTECTED] Homepage: http://jmmv.cjb.net
Re: corruption during power loss
Nate Amsden wrote: Michael Soulier wrote: My biggest complaint about RedHat/Mandrake while I was using them was the fact that if I lost power, the disk caching would cause the filesystem to be corrupted, often seriously so. I'd cringe when I booted up again, because inevitably, I'd be prompted to login as root and run fsck myself. Often I'd have to reinstall a bunch of packages, and it didn't help that doing an rpm -Va to verify everything inevitably returned false negatives on package integrity due to broken rpms. Then I installed Debian. I've had about 5 losses of power since, and each time, the system has come up without a scratch. Minor inode problems easily fixed by fsck without manual control necessary. I noticed the entries in /etc/inittab for powerloss, but the script it's pointing to for me is not installed, so it's not that, although I'd like to know what this /etc/init.d/powerfail script is. So I ask, what aspect of Debian makes it superior in this respect? What's causing the wonderful lack of corruption during power failures? i'd highly reccomend getting a decent UPS. they are quite cheap, and they will not only prevent curroption during outages but it will make the system more stable(its usually difficult/impossible to notice spikes/brownouts by the time the computer shows signs it may be too late..) the life of the hardware will be improved, and a lot of headaches will go away. i can't stress enough a UPS to anyone that is even slightly serious about computers no matter what OS or filesystem they run, the risk to damaging hardware(IMO) is greater then the risk to damaging software. you are in canada so you may not be affected by the serious power problems the U.S. is currently experiencing(wide spread black/brown outs). because of the deregulation of the power industry here there hasn't been a new power plant built in more then 5 years, infact many have closed down(i can think of at least 1 nuclear power plant that closed where i used to live). the situation is going to get worse before it gets better. UPSs are not just for servers, they should be just as standard as a keyboard or a mouse to a PC. kinda sad that people don't take it seriously enough. and btw UPSs are really more useful for cleaning up dirty power rather then protecting against full blown blackouts. ive seen many instances where APC smart UPSs registered severe voltage drop offs(going from 121V to as low as 50-60VA), may not be enough to kill the lights, but probably enough to kill your PC(s). Of course I do agree that a UPS is more than useful. However there seems to be something in the setting-up of debian that helps it surviving a loss of power. It would be interesting to know what it is, maybe it is something in the configuration of the (root) filesystem, like making it sync ? I can see another circumstance where it would be useful not to have the system to crash like redhat/mandrake do (or if you could reproduce on these what make debian survive) : I work at the university where we have students to use machines on free access, these machines dual-boot linux (redhat or is it mandrake now ? BTNTQÂ :-) and windows. I regret that students use windows more than linux, but I have to reckon that it often happens that linux is asking for admin and windows, even if scandisk-ing, starts. The reason of that is very simple, the students turn the computers off without halting the system (can't get them to understand they have to halt it), and it seems that even windows 95 accepts that better than redhat/mandrake. On the other hand I'm pretty sure I couldn't get the admins to install debian instead, but if I came with a handy solution to prevent crashes, by configuring the filesystem or whatever in their redhat/mandrake, they might consider it, as of course it would be less work for them afterwards :-) -- Eric deplagne AKA DelTree E-mail : [EMAIL PROTECTED]
Re: corruption during power loss
On Thu, Sep 14, 2000 at 10:13:53PM -0800, Ethan Benson wrote: if you want your linux filesystems to be safer and are willing to accept the significant performance hit change defaults to defaults,sync in /etc/fstab for your ext2 filesystems. be prepared for things like tar -x and rm -rf to take eons though. (dpkg upgrades take much longer too) Nice to know that I have the option. So far, as I said, I haven't seen any major problems under Debian. Mike
RE: corruption during power loss
You can also get more information regarding JFS from the following site : http://www2.software.ibm.com/developer/tools.nsf/dw/linux-projects-byname?Op enDocumentCount=500 Go to the site and select JFS for Linux. The FAQ may be of interest to you guys. Cheers, Saranjit Singh. -Original Message- From: Ethan Benson [mailto:[EMAIL PROTECTED] Sent: Saturday, September 16, 2000 8:25 AM To: Debian Users List Subject: Re: corruption during power loss On Fri, Sep 15, 2000 at 10:58:57PM +0200, Sven Burgener wrote: On Thu, Sep 14, 2000 at 10:13:53PM -0800, Ethan Benson wrote: your other option is using a Journeling filesystem such as Reiser or ext3 (reiser i think is more mature at this point but still has some serious limitations such as being unsuitable for use on /) It's time for Linux to integrate a journaling FS. Will this be happening with the upcoming 2.4? I mean 'integrated by default'. probably not, from my readings of linux-kernel it appears that the people in charge (Linus Torvalds and Alan Cox)* do not yet feel Reiser is quite ready to be merged into the main kernel tree. i think Alexander Viro the VFS guy also has some issues. The other problem is Linus is getting more conservative about allowing new stuff into 2.4 since the more that is added the longer it will take to finish. Also, which of the journaling FS projects will be the one integrated into the main kernel source tree (sometime)? most likely Reiser will be first since its a bit further along then ext3. But iirc Alan Cox wants to see the Reiser and ext3 journaling layers merged before either goes in. read linux kernel archives or past issues of kernel-cousin on linuxcare. if your interested in the details. * i don't presume to speak for anyone, this is just what i remember reading in the past couple months. -- Ethan Benson http://www.alaska.net/~erbenson/
Re: corruption during power loss
Pollywog wrote: About half the time that I experience power failures, I need to run fsck on my Debian system. How is your system partitioned? One big root partition is not a good idea for someone who suffers frequent power failures. -- John Hasler [EMAIL PROTECTED] (John Hasler) Dancing Horse Hill Elmwood, WI
Re: corruption during power loss
On 15-Sep-2000 John Hasler wrote: Pollywog wrote: About half the time that I experience power failures, I need to run fsck on my Debian system. How is your system partitioned? One big root partition is not a good idea for someone who suffers frequent power failures. I have one big partition plus a swap partition. I was unable to partition my drive the way I wanted to do it. I don't recall the exact problem I had. -- Andrew
Re: corruption during power loss
On Thu, Sep 14, 2000 at 11:51:01PM -0400, Michael Soulier wrote: So, have you had to reinstall packages? With RedHat/Mandrake, I would immediately do an rpm -Va to verify everything, but I quickly found out that several packages would complain about being broken even after a fresh install, so the information was useless. yeah rpm -V is largely useless. (rpm period is largely useless.. oops flamebait ;p) I'm still getting deeper into Debian's packaging tools. for verification there is debsums but not all packages come with md5 lists. I tried running sync on a cronjob every hour, but it didn't help you would have to run it more often then that, like every 1 or 2 minutes, but that is not really a very good solution. much. I never understood why Linux seemed so vulnerable to this. I've had the system crippled for hours with corrupt files while I reinstall packages. Not since installing Debian, but considering what you've said here, maybe I've been lucky. the reason is linux uses filesystem caching extensivly, this GREATLY improves filesystem speed but at the cost of reliability in the event of power failures. BSD tends to do caching much less and as a result has a bit more reliability but at the cost of the filesystem being slow as snot (comparing an OpenBSD box i have here to my GNU/Linux boxen) BSD mounts the filesystems in sync mode rather then async like linux does. BSD has been working on something in between called soft updates, i tried it on my OpenBSD box and ended up with corrupted filesystems... if you want your linux filesystems to be safer and are willing to accept the significant performance hit change defaults to defaults,sync in /etc/fstab for your ext2 filesystems. be prepared for things like tar -x and rm -rf to take eons though. (dpkg upgrades take much longer too) your other option is using a Journeling filesystem such as Reiser or ext3 (reiser i think is more mature at this point but still has some serious limitations such as being unsuitable for use on /) everything is a tradeoff. -- Ethan Benson http://www.alaska.net/~erbenson/ pgpAvWnzyF06P.pgp Description: PGP signature
Re: corruption during power loss
On Thu, Sep 14, 2000 at 10:13:53PM -0800, Ethan Benson wrote: your other option is using a Journeling filesystem such as Reiser or ext3 (reiser i think is more mature at this point but still has some serious limitations such as being unsuitable for use on /) It's time for Linux to integrate a journaling FS. Will this be happening with the upcoming 2.4? I mean 'integrated by default'. Also, which of the journaling FS projects will be the one integrated into the main kernel source tree (sometime)? Sven -- We will run this with the same kind of openness we have run Windows, Steve Ballmer on their .net service
Re: corruption during power loss
On Fri, Sep 15, 2000 at 10:58:57PM +0200, Sven Burgener wrote: On Thu, Sep 14, 2000 at 10:13:53PM -0800, Ethan Benson wrote: your other option is using a Journeling filesystem such as Reiser or ext3 (reiser i think is more mature at this point but still has some serious limitations such as being unsuitable for use on /) It's time for Linux to integrate a journaling FS. Will this be happening with the upcoming 2.4? I mean 'integrated by default'. probably not, from my readings of linux-kernel it appears that the people in charge (Linus Torvalds and Alan Cox)* do not yet feel Reiser is quite ready to be merged into the main kernel tree. i think Alexander Viro the VFS guy also has some issues. The other problem is Linus is getting more conservative about allowing new stuff into 2.4 since the more that is added the longer it will take to finish. Also, which of the journaling FS projects will be the one integrated into the main kernel source tree (sometime)? most likely Reiser will be first since its a bit further along then ext3. But iirc Alan Cox wants to see the Reiser and ext3 journaling layers merged before either goes in. read linux kernel archives or past issues of kernel-cousin on linuxcare. if your interested in the details. * i don't presume to speak for anyone, this is just what i remember reading in the past couple months. -- Ethan Benson http://www.alaska.net/~erbenson/ pgpeSA0wNiXVh.pgp Description: PGP signature
corruption during power loss
My biggest complaint about RedHat/Mandrake while I was using them was the fact that if I lost power, the disk caching would cause the filesystem to be corrupted, often seriously so. I'd cringe when I booted up again, because inevitably, I'd be prompted to login as root and run fsck myself. Often I'd have to reinstall a bunch of packages, and it didn't help that doing an rpm -Va to verify everything inevitably returned false negatives on package integrity due to broken rpms. Then I installed Debian. I've had about 5 losses of power since, and each time, the system has come up without a scratch. Minor inode problems easily fixed by fsck without manual control necessary. I noticed the entries in /etc/inittab for powerloss, but the script it's pointing to for me is not installed, so it's not that, although I'd like to know what this /etc/init.d/powerfail script is. So I ask, what aspect of Debian makes it superior in this respect? What's causing the wonderful lack of corruption during power failures? Thanks guys, Mike To listen to the words of the learned, and to instill into others the lessons of science, is better than religious exercises. -- Prophet Muhammad (pbuh)
Re: corruption during power loss
Huh, that's weird. I used RH more than a little -- my home system had RH (5.2 and 6.0) over a year before it was graced with Debian, and I've used RH 6.1 and 6.2 at work. Though there are bookoos of things I like better about Debian, I never had the problem you described below. I had the power go out on the system at least once with the 6.1 and 6.2 versions, and 5-10 times with the older versions. (The power supply at work is whack, and back then I used to enjoy trying to screw up my system somewhy ;) In every single case, fsck fixed the problems on reboot, just like on debian. Objections: 1) My statistics aren't superb. Especially on recent distros, which Mike is probably talking about. 2) I've not dealt with Mandrake. Aaron On Thu, 14 Sep 2000, Michael Soulier wrote: So I ask, what aspect of Debian makes it superior in this respect? What's causing the wonderful lack of corruption during power failures? Thanks guys, Mike
Re: corruption during power loss
Michael Soulier wrote: snip problems easily fixed by fsck without manual control necessary. I noticed the entries in /etc/inittab for powerloss, but the script it's pointing to for me is not installed, so it's not that, although I'd like to know what this /etc/init.d/powerfail script is. IIRC that's for those with a UPS connected. The UPS can notify the system that the power has failed - usually via the serial port, I believe. Once the system has been notified, the system can respond by running the script pointed to in the powerfail entry in the /etc/inittab There's probably a lot more detail to it than that, but that's the basics. There's some more information in the man page for inittab. -- Mike Werner KA8YSD | He that is slow to believe anything and | everything is of great understanding, '91 GS500E| for belief in one false principle is the Morgantown WV | beginning of all unwisdom. pgp6VfXE51lLV.pgp Description: PGP signature
RE: corruption during power loss
About half the time that I experience power failures, I need to run fsck on my Debian system. On 15-Sep-2000 Michael Soulier wrote: So I ask, what aspect of Debian makes it superior in this respect? What's causing the wonderful lack of corruption during power failures?
RE: corruption during power loss
On Fri, 15 Sep 2000, Pollywog wrote: About half the time that I experience power failures, I need to run fsck on my Debian system. So, have you had to reinstall packages? With RedHat/Mandrake, I would immediately do an rpm -Va to verify everything, but I quickly found out that several packages would complain about being broken even after a fresh install, so the information was useless. I'm still getting deeper into Debian's packaging tools. I tried running sync on a cronjob every hour, but it didn't help much. I never understood why Linux seemed so vulnerable to this. I've had the system crippled for hours with corrupt files while I reinstall packages. Not since installing Debian, but considering what you've said here, maybe I've been lucky. Mike
RE: corruption during power loss
On 15-Sep-2000 Michael Soulier wrote: On Fri, 15 Sep 2000, Pollywog wrote: About half the time that I experience power failures, I need to run fsck on my Debian system. So, have you had to reinstall packages? With RedHat/Mandrake, I would immediately do an rpm -Va to verify everything, but I quickly found out that several packages would complain about being broken even after a fresh install, so the information was useless. I had to reinstall one or two packages once, including inetd, because /etc/inetd.conf had gotten corrupted. Other than that, I have not had to install packages after a power failure. I need to get a UPS. -- Andrew
