Re: Sunday 13th of October: SSD cache test day
On 10/15/2013 09:13 PM, Rolf Fokkens wrote: On 10/14/2013 10:08 AM, Florian Weimer wrote: Is there a write-up somewhere documenting what strategies are implemented by bcache to keep the SSD and the hard disk contents in sync even in the event of a sudden power loss? This is good place to start: http://bcache.evilpiepirate.org/ It doesn't actually address this as far as I can see. It only describes how the data structure integrity is maintained on the SSD side. Even that doesn't seem to address torn writes (which are problem for cheap-to-medium-grade SSDs). The code contains "propagate barriers" as a to-do item (see the beginning of drivers/md/bcache/btree.c). I couldn't find anything that discusses write ordering issues which can occur in write-through mode. -- Florian Weimer / Red Hat Product Security Team -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On Oct 14, 2013, at 2:08 AM, Florian Weimer wrote: > On 10/10/2013 12:02 AM, Rolf Fokkens wrote: > >> The Fedora SSD Cache is this sunday October 13th 2013. This Fedora Test >> Day will focus on bcache based SSD Caching in Fedora 20. >> >> https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache >> >> If you're interested in trying out the new bcache SSD caching >> functionality step by step instructions are available for: > > Is there a write-up somewhere documenting what strategies are implemented by > bcache to keep the SSD and the hard disk contents in sync even in the event > of a sudden power loss? I'd check lkml. The conversations go back a few years. But I think it's good to know that getting your data to stable storage means getting it to either the SSD (cache) or HDD (backing device). The migration from cache to backing is COW, so even if that were interrupted it doesn't obliterate the file system. Even for clean shutdowns the cache is dirty, and bcache therefore doesn't really distinguish between clean and unclean shutdowns - so the recovery code is always being exercised. Because of COW and how much faster SSD's commit, overall I'd expect less potential for fs corruption in the face of unclean shutdowns, compared to a non-cached hard driveā¦ assuming there are no bugs that make this supposition wrong. Chris Murphy -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On 10/14/2013 10:08 AM, Florian Weimer wrote: Is there a write-up somewhere documenting what strategies are implemented by bcache to keep the SSD and the hard disk contents in sync even in the event of a sudden power loss? This is good place to start: http://bcache.evilpiepirate.org/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On 10/10/2013 12:02 AM, Rolf Fokkens wrote: The Fedora SSD Cache is this sunday October 13th 2013. This Fedora Test Day will focus on bcache based SSD Caching in Fedora 20. https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache If you're interested in trying out the new bcache SSD caching functionality step by step instructions are available for: Is there a write-up somewhere documenting what strategies are implemented by bcache to keep the SSD and the hard disk contents in sync even in the event of a sudden power loss? -- Florian Weimer / Red Hat Product Security Team -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On 10/12/2013 10:58 PM, Reartes Guillermo wrote: Maybe that can be enhanced to say "cset.uuid" instead of just "uuid" / "set uuid"? (for which i confused it with dev.uuid shown by blkid, since i never used bcache before). cset.uuid can be obtained from the output of "bcache-show-super". Cheers. Thanks, I updated the text! Rolf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache >> Note the set uuid and attach /dev/sdb1 to /dev/sda2: echo > >> /sys/block/bcache0/bcache/attach Maybe that can be enhanced to say "cset.uuid" instead of just "uuid" / "set uuid"? (for which i confused it with dev.uuid shown by blkid, since i never used bcache before). cset.uuid can be obtained from the output of "bcache-show-super". Cheers. On Fri, Oct 11, 2013 at 6:46 PM, Igor Gnatenko wrote: > On Thu, 2013-10-10 at 00:02 +0200, Rolf Fokkens wrote: >> Hi All, >> >> The Fedora SSD Cache is this sunday October 13th 2013. This Fedora Test >> Day will focus on bcache based SSD Caching in Fedora 20. >> >> https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache >> >> If you're interested in trying out the new bcache SSD caching >> functionality step by step instructions are available for: >> >> - bcache on physical hardware >> - bcache in a virtual machine >> - non-root FS on bcache (with or without LVM) >> - root FS on bcache (wtih or without LVM) >> >> The objective of this Test day is to demonstrate a working Fedora 20 >> system using bcache. Te be more specific: >> >> * The system boots OK; after booting bcache is operating as expected >> * The system updates ("yum update") OK. After updating specifically >> the kernel the system boots OK. >> * The system is bootable when the caching device is disabled. >> >> Although testing on real hardware is closest to "the real thing", >> testing in a VM may also provide good insights on the proper working of >> bcache (except for performance). >> >> If you can't make the date of the test day, adding test case results to the >> wiki anytime next week is fine as well. Though if you do plan on showing up >> to the test day, >> please add your name to the participant list on the wiki, and when the day >> arrives, pop into #fedora-test-day on freenode and give us a shout! If you >> can't make the date >> of the test day, adding test case results to the wiki anytime next week is >> fine as well. Though if you do plan on showing up to the test day, add your >> name to the >> participant list on the wiki, and when the day arrives, pop into >> #fedora-test-day on freenode and give us a shout! >> >> The Wiki page is still under development, so expect some improvements >> before sunday. >> >> Thanks, >> >> Igor Gnatenko >> Rolf Fokkens >> > > Today I've updated wiki page. > At test day will be Kent Overstreet (py1hon) which author of bcache. > > -- > Igor Gnatenko > Fedora release 20 (Heisenbug) > Linux 3.11.4-301.fc20.x86_64 > > -- > devel mailing list > devel@lists.fedoraproject.org > https://admin.fedoraproject.org/mailman/listinfo/devel > Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Sunday 13th of October: SSD cache test day
On Thu, 2013-10-10 at 00:02 +0200, Rolf Fokkens wrote: > Hi All, > > The Fedora SSD Cache is this sunday October 13th 2013. This Fedora Test > Day will focus on bcache based SSD Caching in Fedora 20. > > https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache > > If you're interested in trying out the new bcache SSD caching > functionality step by step instructions are available for: > > - bcache on physical hardware > - bcache in a virtual machine > - non-root FS on bcache (with or without LVM) > - root FS on bcache (wtih or without LVM) > > The objective of this Test day is to demonstrate a working Fedora 20 > system using bcache. Te be more specific: > > * The system boots OK; after booting bcache is operating as expected > * The system updates ("yum update") OK. After updating specifically > the kernel the system boots OK. > * The system is bootable when the caching device is disabled. > > Although testing on real hardware is closest to "the real thing", > testing in a VM may also provide good insights on the proper working of > bcache (except for performance). > > If you can't make the date of the test day, adding test case results to the > wiki anytime next week is fine as well. Though if you do plan on showing up > to the test day, > please add your name to the participant list on the wiki, and when the day > arrives, pop into #fedora-test-day on freenode and give us a shout! If you > can't make the date > of the test day, adding test case results to the wiki anytime next week is > fine as well. Though if you do plan on showing up to the test day, add your > name to the > participant list on the wiki, and when the day arrives, pop into > #fedora-test-day on freenode and give us a shout! > > The Wiki page is still under development, so expect some improvements > before sunday. > > Thanks, > > Igor Gnatenko > Rolf Fokkens > Today I've updated wiki page. At test day will be Kent Overstreet (py1hon) which author of bcache. -- Igor Gnatenko Fedora release 20 (Heisenbug) Linux 3.11.4-301.fc20.x86_64 -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Sunday 13th of October: SSD cache test day
Hi All, The Fedora SSD Cache is this sunday October 13th 2013. This Fedora Test Day will focus on bcache based SSD Caching in Fedora 20. https://fedoraproject.org/wiki/Test_Day:2013-10-13_SSD_Cache If you're interested in trying out the new bcache SSD caching functionality step by step instructions are available for: - bcache on physical hardware - bcache in a virtual machine - non-root FS on bcache (with or without LVM) - root FS on bcache (wtih or without LVM) The objective of this Test day is to demonstrate a working Fedora 20 system using bcache. Te be more specific: * The system boots OK; after booting bcache is operating as expected * The system updates ("yum update") OK. After updating specifically the kernel the system boots OK. * The system is bootable when the caching device is disabled. Although testing on real hardware is closest to "the real thing", testing in a VM may also provide good insights on the proper working of bcache (except for performance). If you can't make the date of the test day, adding test case results to the wiki anytime next week is fine as well. Though if you do plan on showing up to the test day, please add your name to the participant list on the wiki, and when the day arrives, pop into #fedora-test-day on freenode and give us a shout! If you can't make the date of the test day, adding test case results to the wiki anytime next week is fine as well. Though if you do plan on showing up to the test day, add your name to the participant list on the wiki, and when the day arrives, pop into #fedora-test-day on freenode and give us a shout! The Wiki page is still under development, so expect some improvements before sunday. Thanks, Igor Gnatenko Rolf Fokkens -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct