Re: [e-users] Crash on e5
On 09/12/2013 08:40, Nex6 wrote: * Carsten Haitzler ras...@rasterman.com [2013-12-08 21:42:25 +0900]: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. I am going to chime in and give my 2 cents. for filesystems, on production servers I tell our ops guys only use ext4 or xfs thats it. Gentoo on critical production server that face the public without a proper QA staging setup? Good lord, that's a scary thought. I banned gentoo from production around here. Lest anyone think I'm a gentoo-hater, my personal workstation: $ uname -a Linux khamul 3.12.3-gentoo #1 SMP PREEMPT Sun Dec 8 00:13:02 SAST 2013 ^ -- Alan McKinnon alan.mckin...@gmail.com -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Mon, 09 Dec 2013 10:01:14 +0200 Alan McKinnon alan.mckin...@gmail.com wrote: On 09/12/2013 08:40, Nex6 wrote: * Carsten Haitzler ras...@rasterman.com [2013-12-08 21:42:25 +0900]: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. I am going to chime in and give my 2 cents. for filesystems, on production servers I tell our ops guys only use ext4 or xfs thats it. Gentoo on critical production server that face the public without a proper QA staging setup? Good lord, that's a scary thought. I banned gentoo from production around here. Lest anyone think I'm a gentoo-hater, my personal workstation: $ uname -a Linux khamul 3.12.3-gentoo #1 SMP PREEMPT Sun Dec 8 00:13:02 SAST 2013 ^ Meh we all have our different beliefs about such things. I personally use BTRFS at home, and switched from ZFS to BTRFS on a production server for a client. This thread can devolve into a my distro / FS is better than yours war, or we can just let the person that volunteered to do the work do it the way he's most comfortable with. -- A big old stinking pile of genius that no one wants coz there are too many silver coated monkeys in the world. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On 07/12/2013 22:16, Bertrand Jacquin wrote: On 2013-12-07 22:09, Guillaume Friloux wrote: On 07/12/2013 21:05, Bertrand Jacquin wrote: Moving to other FS is not an option. BTRFS that format is not yet completed/fully defined, has elementary fsck when not giving needed features. ZFS using FUSE is not going to happened on that critical host. Kernel was a 3.7 and have been updated to longterm-stable 3.10 kernel. There are been a lot of fixes on EXT and VFS between 3.7 and 3.10. We stick on longterm kernel on all hosts, that one was the last not using a longterm stable. Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. I dont say we need it. It was only for information. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. [1] http://zfsonlinux.org/ -- Regards, Mick -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sun, 8 Dec 2013 21:42:25 +0900 Carsten Haitzler (The Rasterman) ras...@rasterman.com wrote: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. Absolutely. The other filesystems are amazing in their feature sets, but are not viable production filesystems quite yet in my opinion. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. Gentoo is a great desktop distro, but definitely not a server OS (may cause a flame war here, but sorry...). For me, and I've been a sysadmin since Debian first came out (yes, that long), and I've used literally all distros at one time or another. Debian stable (or even testing) is an ideal server OS you can absolutely rely on. I would not use anything else. If you absolutely must use gentoo for some feature only it provides, then only use it in the VM - definitely NOT on the host. The host must be rock solid. But, if rebuilding the setup is on the radar in light of this wake up call, the filesystem argument is totally moot - just use LVM for the images. It's the correct thing to do for a production VM host. -- Regards, Christopher Barry Random geeky fortune: One does not thank logic. -- Sarek, Journey to Babel, stardate 3842.4 -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sun, 8 Dec 2013 12:17:53 -0500 Christopher Barry christopher.r.ba...@gmail.com said: On Sun, 8 Dec 2013 21:42:25 +0900 Carsten Haitzler (The Rasterman) ras...@rasterman.com wrote: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. Absolutely. The other filesystems are amazing in their feature sets, but are not viable production filesystems quite yet in my opinion. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. Gentoo is a great desktop distro, but definitely not a server OS (may cause a flame war here, but sorry...). For me, and I've been a sysadmin since Debian first came out (yes, that long), and I've used literally all distros at one time or another. Debian stable (or even testing) is an ideal server OS you can absolutely rely on. I would not use anything else. If you absolutely must use gentoo for some feature only it provides, then only use it in the VM - definitely NOT on the host. The host must be rock solid. But, if rebuilding the setup is on the radar in light of this wake up call, the filesystem argument is totally moot - just use LVM for the images. It's the correct thing to do for a production VM host. well beber (our admin who volunteers his time to the server and its vms) wants gentoo. -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On 2013-12-09 01:52, Carsten Haitzler wrote: On Sun, 8 Dec 2013 12:17:53 -0500 Christopher Barry christopher.r.ba...@gmail.com said: On Sun, 8 Dec 2013 21:42:25 +0900 Carsten Haitzler (The Rasterman) ras...@rasterman.com wrote: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. Absolutely. The other filesystems are amazing in their feature sets, but are not viable production filesystems quite yet in my opinion. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. Gentoo is a great desktop distro, but definitely not a server OS (may cause a flame war here, but sorry...). For me, and I've been a sysadmin since Debian first came out (yes, that long), and I've used literally all distros at one time or another. Debian stable (or even testing) is an ideal server OS you can absolutely rely on. I would not use anything else. If you absolutely must use gentoo for some feature only it provides, then only use it in the VM - definitely NOT on the host. The host must be rock solid. But, if rebuilding the setup is on the radar in light of this wake up call, the filesystem argument is totally moot - just use LVM for the images. It's the correct thing to do for a production VM host. well beber (our admin who volunteers his time to the server and its vms) wants gentoo. Yep :) And opposed to Christopher, I don't find gentoo is great for a desktop but for servers. The main reason is flexibility and that it's easier to manage, but as Chris said, this is flamewar topic, I could write a book on why people should not use Debian on servers. We have a builder host, so configuration, packages etc are shared across all servers and no compilation is done on production hosts (without that, I have to say that it's a real pain in ass to maintain, but not the case) About FS, raster speech agreed . -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
* Carsten Haitzler ras...@rasterman.com [2013-12-08 21:42:25 +0900]: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. I am going to chime in and give my 2 cents. for filesystems, on production servers I tell our ops guys only use ext4 or xfs thats it. Nex6 -- - Codito, ergo sum - I code, therefore I am -- The Rasterman (Carsten Haitzler)ras...@rasterman.com -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
* Bertrand Jacquin be...@meleeweb.net [2013-12-09 02:41:18 +0100]: On 2013-12-09 01:52, Carsten Haitzler wrote: On Sun, 8 Dec 2013 12:17:53 -0500 Christopher Barry christopher.r.ba...@gmail.com said: On Sun, 8 Dec 2013 21:42:25 +0900 Carsten Haitzler (The Rasterman) ras...@rasterman.com wrote: On Sun, 8 Dec 2013 10:25:04 + Mick michaelkintz...@gmail.com said: On Sunday 08 Dec 2013 01:48:18 Steven@e wrote: Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. There are out-of-tree Linux kernel modules care of ZFSOnLinux Project[1]. Therefore using ZFS-Fuse is not necessary (or recommended). At least one valid reason for using ZFS (there are many) is that it guards against fs corruption by using CRC checksums. I understand that both Oracle (RHL) Linux and SUSE consider BTRFS production ready and Oracle will be/are using this instead of ZFS. From my limited understanding BTRFS is being developed at speed and catching up with ZFS, but it does not have the amount of testing that ZFS had to date to vouch for its stability/maturity. At this stage in their development ZFS is superior to BTRFS in terms of functionality, although there is hope that BTRFS will develop at speed. and why? ext4 HAS been production ready for YEARS... inf act not production ready... it has been *IN8 production for years... if there is a fs i would trust - it's ext4. not zfs and DEFINITELY not btrfs. ext4 (and 3 etc. before it) have many more miles of PRODUCTION behind them. Absolutely. The other filesystems are amazing in their feature sets, but are not viable production filesystems quite yet in my opinion. what this probably was ... was an unstable bleeding-edge kernel since the servers are being run on gentoo and thus are not exactly being conservative. it was probably a newly introduced bug that hasn't been hammered out and other fs's used less will have such bugs many times MORE than ext4 will. Gentoo is a great desktop distro, but definitely not a server OS (may cause a flame war here, but sorry...). For me, and I've been a sysadmin since Debian first came out (yes, that long), and I've used literally all distros at one time or another. Debian stable (or even testing) is an ideal server OS you can absolutely rely on. I would not use anything else. If you absolutely must use gentoo for some feature only it provides, then only use it in the VM - definitely NOT on the host. The host must be rock solid. But, if rebuilding the setup is on the radar in light of this wake up call, the filesystem argument is totally moot - just use LVM for the images. It's the correct thing to do for a production VM host. well beber (our admin who volunteers his time to the server and its vms) wants gentoo. Yep :) And opposed to Christopher, I don't find gentoo is great for a desktop but for servers. The main reason is flexibility and that it's easier to manage, but as Chris said, this is flamewar topic, I could write a book on why people should not use Debian on servers. We have a builder host, so configuration, packages etc are shared across all servers and no compilation is done on production hosts (without that, I have to say that it's a real pain in ass to maintain, but not the case) About FS, raster speech agreed . while I love using debian for servers or desktops i think using the distro your most familer with is best. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
[e-users] Crash on e5
Hi, This morning e5 have been unavailable for 2 hours and 30 minutes starting at about 10:40 UTC. All is back again, but if you have strange behaviour, please reply (and CC: me). Thanks to thoses who already did. The EXT4 filesystem that host the VMs used for Enlightenment internal and public purposes are stored on a RADI10 partition that protect us against a disc failure but not against a filesystem corruption what happened to us. I don't known exactly the source of that corruption but in any case it took some time to fsck, and verify data integrity. Corrupted file system: - e5: /data - VMs storage Everything OK after a long fsck - mail1: /srv/mail - mail storage Missing subscribers on git list, e-b0rk list entirely corrupted Backup restored in date of 06/12/2013 - web1: /srv/mysql - MySQL storage Databases corrupted, binlog corrupted, complete FS corrupted Restored a version of mid 2013, backup are missing. I need to fix my backup configuration on this. The only use and so lost database is the one used by exchange.enlightenment.org what lead to addition made since mid 2013 are lost. - build-gentoo-cross1: /srv/build - compile storage Corrupted source and compiled file Not restored as not backuped, can be regenerated, all history is present on jenkins-master1 All VMs FS have been checked, as I say everything should be OK, but report any abnormal things. A chance we have outsourced backup. -- Beber -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sat, 7 Dec 2013 17:41:36 +0100 Bertrand Jacquin be...@meleeweb.net wrote: Missing subscribers on git list, e-b0rk list entirely corrupted e-b0rk got borked, think that was just asking for trouble. B-) -- A big old stinking pile of genius that no one wants coz there are too many silver coated monkeys in the world. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Saturday 07 Dec 2013 17:22:06 David Seikel wrote: On Sat, 7 Dec 2013 17:41:36 +0100 Bertrand Jacquin be...@meleeweb.net wrote: Missing subscribers on git list, e-b0rk list entirely corrupted e-b0rk got borked, think that was just asking for trouble. B-) Perhaps time to move to ZFS? -- Regards, Mick -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sat, 7 Dec 2013 18:11:16 + Mick michaelkintz...@gmail.com wrote: On Saturday 07 Dec 2013 17:22:06 David Seikel wrote: On Sat, 7 Dec 2013 17:41:36 +0100 Bertrand Jacquin be...@meleeweb.net wrote: Missing subscribers on git list, e-b0rk list entirely corrupted e-b0rk got borked, think that was just asking for trouble. B-) Perhaps time to move to ZFS? Or BTRFS? -- A big old stinking pile of genius that no one wants coz there are too many silver coated monkeys in the world. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On Sun, 8 Dec 2013 04:20:37 +1000 David Seikel onef...@gmail.com wrote: On Sat, 7 Dec 2013 18:11:16 + Mick michaelkintz...@gmail.com wrote: On Saturday 07 Dec 2013 17:22:06 David Seikel wrote: On Sat, 7 Dec 2013 17:41:36 +0100 Bertrand Jacquin be...@meleeweb.net wrote: Missing subscribers on git list, e-b0rk list entirely corrupted e-b0rk got borked, think that was just asking for trouble. B-) Perhaps time to move to ZFS? Or BTRFS? Regardless of filesystem type, determining the cause of the corruption should be attempted and corrected first if possible. Could multiple VMs somehow be writing to the same data in the main FS simultaneously per chance? (just a thought, as this is often a cause of corruption in non-clustered filesystems, but I do not fully understand your config) Also, running the VMs as disk files in a filesystem should really be avoided if at all possible. In my experience, it's better to use logical volumes directly for VM images for performance, scalability, and generally less layers of stuff to fail in the IO path. And of course, a simple nightly rsync of stuff that matters should be setup... but you know that quite well now :( I've had very good experiences with this: https://github.com/DrHyde/rsnapshot It's simple, and versions the backups with excellent space, time and bandwidth utilization so you can roll back to configurable points in time easily, without consuming a bunch of space unnecessarily. -- Regards, Christopher Barry Random geeky fortune: Dump the condiments. If we are to be eaten, we don't need to taste good. -- Visionaries cartoon -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On 2013-12-07 20:30, Christopher Barry wrote: On Sun, 8 Dec 2013 04:20:37 +1000 David Seikel onef...@gmail.com wrote: On Sat, 7 Dec 2013 18:11:16 + Mick michaelkintz...@gmail.com wrote: On Saturday 07 Dec 2013 17:22:06 David Seikel wrote: On Sat, 7 Dec 2013 17:41:36 +0100 Bertrand Jacquin be...@meleeweb.net wrote: Missing subscribers on git list, e-b0rk list entirely corrupted e-b0rk got borked, think that was just asking for trouble. B-) Perhaps time to move to ZFS? Or BTRFS? Moving to other FS is not an option. BTRFS that format is not yet completed/fully defined, has elementary fsck when not giving needed features. ZFS using FUSE is not going to happened on that critical host. Kernel was a 3.7 and have been updated to longterm-stable 3.10 kernel. There are been a lot of fixes on EXT and VFS between 3.7 and 3.10. We stick on longterm kernel on all hosts, that one was the last not using a longterm stable. Regardless of filesystem type, determining the cause of the corruption should be attempted and corrected first if possible. Could multiple VMs somehow be writing to the same data in the main FS simultaneously per chance? (just a thought, as this is often a cause of corruption in non-clustered filesystems, but I do not fully understand your config) Nop :) Also, running the VMs as disk files in a filesystem should really be avoided if at all possible. In my experience, it's better to use logical volumes directly for VM images for performance, scalability, and generally less layers of stuff to fail in the IO path. In my experience too and it was designed like this at the beginning before moving to qcow2 files argued by others people that it make easier administration, which what I still don't agree and add another layer of abstraction (with more cache, and sync issues). And of course, a simple nightly rsync of stuff that matters should be setup... but you know that quite well now :( That's what is done using other tools than rsnapshot. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On 07/12/2013 21:05, Bertrand Jacquin wrote: Moving to other FS is not an option. BTRFS that format is not yet completed/fully defined, has elementary fsck when not giving needed features. ZFS using FUSE is not going to happened on that critical host. Kernel was a 3.7 and have been updated to longterm-stable 3.10 kernel. There are been a lot of fixes on EXT and VFS between 3.7 and 3.10. We stick on longterm kernel on all hosts, that one was the last not using a longterm stable. Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
On 2013-12-07 22:09, Guillaume Friloux wrote: On 07/12/2013 21:05, Bertrand Jacquin wrote: Moving to other FS is not an option. BTRFS that format is not yet completed/fully defined, has elementary fsck when not giving needed features. ZFS using FUSE is not going to happened on that critical host. Kernel was a 3.7 and have been updated to longterm-stable 3.10 kernel. There are been a lot of fixes on EXT and VFS between 3.7 and 3.10. We stick on longterm kernel on all hosts, that one was the last not using a longterm stable. Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
Re: [e-users] Crash on e5
ZFS on linux is clearly not stable yet. not an option. On Sat, Dec 7, 2013 at 10:16 PM, Bertrand Jacquin be...@meleeweb.net wrote: On 2013-12-07 22:09, Guillaume Friloux wrote: On 07/12/2013 21:05, Bertrand Jacquin wrote: Moving to other FS is not an option. BTRFS that format is not yet completed/fully defined, has elementary fsck when not giving needed features. ZFS using FUSE is not going to happened on that critical host. Kernel was a 3.7 and have been updated to longterm-stable 3.10 kernel. There are been a lot of fixes on EXT and VFS between 3.7 and 3.10. We stick on longterm kernel on all hosts, that one was the last not using a longterm stable. Hello beber, just for information, you dont need FUSE anymore to use ZFS. zfsonlinux solves this. Thanks for the info, but why should we need this ? I see no valid reason. -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users -- Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631iu=/4140/ostg.clktrk ___ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users