Re: Checking filesystems periodically
On Fri, May 4, 2012 at 9:41 AM, Daniel Drake d...@laptop.org wrote: Thirdly, fsck is not magic. It cannot detect/repair all corruption. As far as I know, we have not yet found a case of corruption which can be meaningfully fixed by fsck. We did do quite a bit of testing for this at an earlier point. It is worthwhile expanding on this: ext3 is a journalled fs, so it can fix _some_ issues thanks to the journal, and does so on boot. So the harsh environment issues are mostly handled by ext3 journal-based recovery. For problems not fixable with ext3's journal-based recovery... cannot (currently) be fixed by fsck as dsd writes. At least that's what we found so far. Given those findings, it fell in our priorities list, and there is harsh competition there! :-) Now, fsck and our choice of FS are not frozen in stone, so help is welcome on this track. Improving our plymouth screens to have an I'm doing extra work this time, so booting slower image would be good. Not easy, but definitely good. (Note that any boot time repair must be fully automated. 6 year olds won't be telling fsck what to do with the broken inode table.) cheers, m -- martin.langh...@gmail.com mar...@laptop.org -- Software Architect - OLPC - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Checking filesystems periodically
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Hi, I am curious to know why we are not periodically checking file systems after every N boots on the XO laptop. The biggest reason I can assume is because every Nth time the system will appear to boot very slowly to the user thus creating the impression that something is wrong. However, in light of the fact that XO's often operate in harsh environments and (probably) greater possibility of hardware failure because of that, would it be a good idea in doing so? Perhaps we could mount /boot and /home on different partitions and check /boot periodically. One issue that is pertinent to this discussion is [1] Thoughts? [1] http://dev.laptop.org.au/issues/1214 Cheers, Anish -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQEcBAEBAgAGBQJPo5mRAAoJEBoxUdDHDZVp3dwH/0vOsuR9zxv/u+2wzyMCIXNb 7SKQeevki8oBaZW6S4fVm5UZUT02hNothrFJhfz8XFwYigahVKyPUnYTPmC9kHVu C88azxuDRHcUeRyuHvVZdp/G7Xyd+FUI+DfjnEi/ts8lgRBMFfiz+g/8MljW2rsN 32Bu6/fevujT3khvkFn3LMKSfRJtVCXMt9PddRHBVXxmIjxS/3JV8kAYJiheCE7o Na3zsTN3G+nTLEXd+oPJ5JeDiJnuz94genQu160vD4nqhtwsDD5HtsTIXf0tp3BD yd8TYSGDHuolSU0UqDfPy52mYuA0qTLBjbBQ7TXnwuUXaGJpTyF+XYFgSaorFeo= =Kl8n -END PGP SIGNATURE- ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
On Fri, May 4, 2012 at 2:55 AM, Anish Mangal an...@activitycentral.com wrote: Hi, I am curious to know why we are not periodically checking file systems after every N boots on the XO laptop. Historically, we have used a filesystem without a checker (jffs2). Now that we use ext* on newer laptops, there are a few reasons: Firstly, its non-trivial, due to the design/setup of our initramfs and filesystem. Secondly, the user experience: The biggest reason I can assume is because every Nth time the system will appear to boot very slowly to the user thus creating the impression that something is wrong. Until F17 we haven't had a good way of communicating this via the boot animation. Now we can do that easily but it lacks implementation. Thirdly, fsck is not magic. It cannot detect/repair all corruption. As far as I know, we have not yet found a case of corruption which can be meaningfully fixed by fsck. We did do quite a bit of testing for this at an earlier point. However, in light of the fact that XO's often operate in harsh environments and (probably) greater possibility of hardware failure because of that, would it be a good idea in doing so? Despite all the above, yes, it would be a good idea, it's something we'll hopefully get around to at some point. Your help implementing it is welcome. Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
Hi, On Fri, May 04 2012, Daniel Drake wrote: Until F17 we haven't had a good way of communicating this via the boot animation. Now we can do that easily but it lacks implementation. Thirdly, fsck is not magic. It cannot detect/repair all corruption. As far as I know, we have not yet found a case of corruption which can be meaningfully fixed by fsck. We did do quite a bit of testing for this at an earlier point. Also, our users cannot be expected to understand (or obey) a requirement that they not turn off the machine while it's doing something dangerous: so if powering down half way through fsck leaves the filesystem in a worse state than it was before fsck ran, we probably shouldn't do it at all. - Chris. -- Chris Ball c...@laptop.org http://printf.net/ One Laptop Per Child ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
Hi, On Fri, May 04 2012, Chris Ball wrote: Hi, On Fri, May 04 2012, Daniel Drake wrote: Until F17 we haven't had a good way of communicating this via the boot animation. Now we can do that easily but it lacks implementation. Thirdly, fsck is not magic. It cannot detect/repair all corruption. As far as I know, we have not yet found a case of corruption which can be meaningfully fixed by fsck. We did do quite a bit of testing for this at an earlier point. Also, our users cannot be expected to understand (or obey) a requirement that they not turn off the machine while it's doing something dangerous: so if powering down half way through fsck leaves the filesystem in a worse state than it was before fsck ran, we probably shouldn't do it at all. Another also: sometimes when fsck finds an inconsistency it asks you for the root password, but some of our users don't have the root password, so they might end up in a reboot loop where they can't progress. - Chris. -- Chris Ball c...@laptop.org http://printf.net/ One Laptop Per Child ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri 04 May 2012 07:57:47 PM IST, Chris Ball wrote: Hi, On Fri, May 04 2012, Chris Ball wrote: Hi, On Fri, May 04 2012, Daniel Drake wrote: Until F17 we haven't had a good way of communicating this via the boot animation. Now we can do that easily but it lacks implementation. Thirdly, fsck is not magic. It cannot detect/repair all corruption. As far as I know, we have not yet found a case of corruption which can be meaningfully fixed by fsck. We did do quite a bit of testing for this at an earlier point. Also, our users cannot be expected to understand (or obey) a requirement that they not turn off the machine while it's doing something dangerous: so if powering down half way through fsck leaves the filesystem in a worse state than it was before fsck ran, we probably shouldn't do it at all. This is a valid point. _If_ there is a possibility that fsck will leave the system in a worse state if the laptop is accidentally powered off, this is a bad idea. However, we can probably do a 'read-only' fsck, and provide a notification (perhaps to contact technical support) if it finds problems. That could be a fail-safe way of implementing this. Am I right in assuming (with my limited knowledge in this area) that the fsck-on-boot is by default read-only? Another also: sometimes when fsck finds an inconsistency it asks you for the root password, but some of our users don't have the root password, so they might end up in a reboot loop where they can't progress. As above, a non-intrusive way of doing this would be providing a notification to contact technical support. The biggest caveat then becomes that the problem doesn't happen too often ;-) or else tech support start cursing us! In general, I wouldn't expect users to open a terminal and type commands and passwords to repair their machine. If it has to be implemented, it has to be completely handled by a GUI. - Chris - -- Anish -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQEcBAEBAgAGBQJPo/NaAAoJEBoxUdDHDZVp7XMH/jQOdpdgeY0KPkgjwu7ImnyA 8kxSJcZUXtQ2bVA9ASXBmQH5+4gW/ZWUl+L0zbj/NwtFxef6clWnyeOtGwoTvRDp pBBNrj/5y3NjeshEvZGFCZH0VxjmNS7I77I5b9f9qVsp2y03ZUSgC8E9z6WPet4q fwAfjpgEOW+P3wclVDiWpzDuWFl6rXkiMssETuIIwPWCVEl04Wo7GikNrQrLe1M/ ySh6XSlZJkMv4RexBWzg/MHb9XNNFy+HwHmvWcD6h4wSUC6GjDBmUi14RvmS1Bz6 +a70xhiOA+bp8rsqvkEBzLHJA89g2BRImuLinez8Y2aR9K+rW1TrfzGwjCffdv8= =q8zv -END PGP SIGNATURE- ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
On Fri, May 4, 2012 at 2:55 AM, Anish Mangalan...@activitycentral.com wrote: I am curious to know why we are not periodically checking file systems after every N boots on the XO laptop. I think the question is: Should functions which affect the 'system' be performed automatically, or should they be explicitly invoked ? The currently-existing OLPC precedent is the 'Software updates' facility -- which gets invoked manually to check whether the system's Activities are up-to-date. My suggestion is to package the required software (fsck?) into a new 'Check system' facility - and add an icon for that new facility to the 'My Settings' panel - then the user who invokes it will know why his XO is tied up for a while. [The local XO-laptop distributing organization can publish guidelines as to how often such a 'Check system' facility ought to be used.] mikus ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Checking filesystems periodically
On Fri, May 4, 2012 at 2:46 PM, Mikus Grinbergs mi...@bga.com wrote: On Fri, May 4, 2012 at 2:55 AM, Anish Mangalan...@activitycentral.com wrote: I am curious to know why we are not periodically checking file systems after every N boots on the XO laptop. I think the question is: Should functions which affect the 'system' be performed automatically, or should they be explicitly invoked ? The currently-existing OLPC precedent is the 'Software updates' facility -- which gets invoked manually to check whether the system's Activities are up-to-date. My suggestion is to package the required software (fsck?) into a new 'Check system' facility - and add an icon for that new facility to the 'My Settings' panel - then the user who invokes it will know why his XO is tied up for a while. [The local XO-laptop distributing organization can publish guidelines as to how often such a 'Check system' facility ought to be used.] +1 mikus ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel -- Walter Bender Sugar Labs http://www.sugarlabs.org ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
