[zfs-discuss] Borked zpool is now invulnerable
Hello, while testing some code changes, I managed to fail an assertion while doing a zfs create. My zpool is now invulnerable to destruction. :( bash-3.00# zpool destroy -f test_undo internal error: unexpected error 0 at line 298 of ../common/libzfs_dataset.c bash-3.00# zpool status pool: test_undo state: ONLINE scrub: none requested config: NAMESTATE READ WRITE CKSUM test_undo ONLINE 0 0 0 c0d1s1ONLINE 0 0 0 c0d1s0ONLINE 0 0 0 errors: No known data errors How can I destroy this pool so I can use the disk for a new pool? Thanks! :) -- Regards, Jeremy ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Borked zpool is now invulnerable
Jeremy Teo wrote: How can I destroy this pool so I can use the disk for a new pool? Thanks! :) dd if=/dev/zero of=/dev/dsk/c0d1s1 dd if=/dev/zero of=/dev/dsk/c0d1s0 that should do it. -- Darren J Moffat ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Borked zpool is now invulnerable
On Thu, 2006-05-18 at 22:05 +0800, Jeremy Teo wrote: > My zpool is now invulnerable to destruction. :( Nifty - does that mean your disk is also invulnerable to hardware errors too ? [ as in, your typical superhero who gets endowed with special abilities due to a failed radiation experiment ;-) ] > bash-3.00# zpool destroy -f test_undo > internal error: unexpected error 0 at line 298 of ../common/libzfs_dataset.c I'd suggest zpool exporting all other pools on the system, then reboot into failsafe mode, mount your root partition somewhere (like /tmp/a) then remove /tmp/a/etc/zfs/zpool.cache. Finally, reboot and create a new pool on the disk which has now been made available, and then zpool import the others. Hope that helps (all disclaimers about making backups of anything you care about before doing this sort of thing apply) Anyone know of a simpler way ? cheers, tim -- Tim Foster, Sun Microsystems Inc, Operating Platforms Group Engineering Operationshttp://blogs.sun.com/timf ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Re: Re: zfs snapshot for backup, Quota
Sorry to revive such an old thread.. but I'm struggling here. I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I work for a University, where everyone has a quota. I'd literally have to create > 10K partitions. Is that really your intention? Of course, backups become a huge pain now. Even with the scripted idea below, that's cumbersome for both backups and (especially) restores. Why can't we just have user quotas in zfs? :) Respectfully, -Charlie This message posted from opensolaris.org ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: > Sorry to revive such an old thread.. but I'm struggling here. > > I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I > work for a University, where everyone has a quota. I'd literally have > to create > 10K partitions. Is that really your intention? Yes. You'd group them all under a single filesystem in the hierarchy, allowing you to manage NFS share options, compression, and more from a single control point. > Of course, backups become a huge pain now. below, that's cumbersome > for both backups and (especially) restores. Using traditional tools or ZFS send/receive? We are working on RFEs for recursive snapshots, send, and recv, as well as preserving DSL properties as part of a 'send', which should make backups of large filesystem hierarchies much simpler. > Why can't we just have user quotas in zfs? :) The fact that per-user quotas exist is really a historical artifact. With traditional filesystems, it is (effectively) impossible to have a filesystem per user. The filesystem is a logical administrative control point, allowing you to view usage, control properties, perform backups, take snapshots etc. For home directory servers, you really want to do these operations per-user, so logically you'd want to equate the two (filesystem = user). Per-user quotas (the most common use of quotas, but not the only one) were introduced because multiple users had to share the same filesystem. ZFS quotas are intentionally not associated with a particular user because a) it's the logical extension of "filesystems as control point", b) it's vastly simpler to implement and, most importantly, c) separates implementation from adminsitrative policy. ZFS quotas can be set on filesystems which may represent projects, groups, or any other abstraction, as well as on entire portions of the hierarchy. This allows them to be combined in ways that traditional per-user quotas cannot. Hope that helps, - Eric -- Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Re: Re: zfs snapshot for backup, Quota
> Why can't we just have user quotas in zfs? :) +1 to that. I support a couple environments with group/user quotas that cannot move to ZFS since they serve brain-dead apps that read/write from a single directory. I also agree that using even a few hundred mountpoints is more tedious than using quotas, but I can get used to that...I just won't use df as often. :) This message posted from opensolaris.org ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Re: Re: zfs snapshot for backup, Quota
Eric Schrock wrote: > > On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: >> >> to create > 10K partitions. Is that really your intention? > > > > Yes. You'd group them all under a single filesystem in the hierarchy, > > allowing you to manage NFS share options, compression, and more from a > > single control point. This isn't so bad. I'm going to assume that mounting 10K partitions at boot doesn't take forever. :) >> >> Of course, backups become a huge pain now. ... that's cumbersome >> >> for both backups and (especially) restores. > > > > Using traditional tools or ZFS send/receive? Traditional (amanda). I'm not seeing a way to dump zfs file systems to tape without resorting to 'zfs send' being piped through gtar or something. Even then, the only thing I could restore was an entire file system. (We frequently restore single files for users...) Perhaps, since zfs isn't limited to one snapshot per FS like fssnap is, I should be redesigning everything. It sounds like I should look at using many snapshots, and dumping to tape (each file system, somehow) less frequently. Waiting for S10_U2 now :) > > Hope that helps, > > > > - Eric It does. Thanks! -Charlie This message posted from opensolaris.org ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, 2006-05-18 at 12:12 -0700, Eric Schrock wrote: > On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: > > Sorry to revive such an old thread.. but I'm struggling here. > > > > I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I > > work for a University, where everyone has a quota. I'd literally have > > to create > 10K partitions. Is that really your intention? > > Yes. You'd group them all under a single filesystem in the hierarchy, > allowing you to manage NFS share options, compression, and more from a > single control point. > I'd agree except for backups. If the pools are going to grow beyond a reasonable-to-backup and reasonable-to-restore threshold (measured by the backup window), it would be practical to break it into smaller pools. After all, you'll probably have to restore a pool eventually. If that will take a week, your users won't be very happy with your solution. > > Of course, backups become a huge pain now. below, that's cumbersome > > for both backups and (especially) restores. > > Using traditional tools or ZFS send/receive? We are working on RFEs for > recursive snapshots, send, and recv, as well as preserving DSL > properties as part of a 'send', which should make backups of large > filesystem hierarchies much simpler. > Using EBS or NetBackup, can I get a single file back from tape only through the backup system? That's a big factor for production environments. Also, when users request a restore from tape from offsite backups, they'll usually specify a date range for when the file was 'good'. To accomplish that, you need to use the backup solution to find the requisite file. These 'fishing expeditions' (as I call them) can take a lot of time if direct access isn't available via the backup tool. I believe you're referring in the above to using zfs send/recv for backup to tape. Until the vendors work with zfs send/recv, it's not a viable option for filesystem backups in a production environment. Related to that, does anybody have a timeframe for direct support for ZFS send/recv (or something similar) in NBU or EBS? [ quota explanation deleted for brevity ] > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > ___ > zfs-discuss mailing list > zfs-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, May 18, 2006 at 02:23:55PM -0600, Gregory Shaw wrote: > I'd agree except for backups. If the pools are going to grow beyond a > reasonable-to-backup and reasonable-to-restore threshold (measured by > the backup window), it would be practical to break it into smaller > pools. Speaking of backups, and particularly when we get to recursive ones, I'd like control over the filesystem and snapshot names restored, as well as control over what snapshots should appear, overrides for properties (considering the RFE to have property setting on zfs create). Currently I think one can specify the fs name to restore, yes, I know. Also, the recursive backup output will need a ToC and a tool to list it. Nico -- ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, May 18, 2006 at 12:46:28PM -0700, Charlie wrote: > Traditional (amanda). I'm not seeing a way to dump zfs file systems to > tape without resorting to 'zfs send' being piped through gtar or > something. Even then, the only thing I could restore was an entire file > system. (We frequently restore single files for users...) > > Perhaps, since zfs isn't limited to one snapshot per FS like fssnap is, > I should be redesigning everything. It sounds like I should look at > using many snapshots, and dumping to tape (each file system, somehow) > less frequently. That's right. With ZFS, there should never be a need to go to tape to recover an accidentally deleted file, becuase it's easy[*] to keep lots of snapshots around. [*] Well, modulo 6373978 "want to take lots of snapshots quickly ('zfs snapshot -r')". I'm working on that... --matt ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On 5/18/06, Gregory Shaw <[EMAIL PROTECTED]> wrote: On Thu, 2006-05-18 at 12:12 -0700, Eric Schrock wrote: > On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: > > Sorry to revive such an old thread.. but I'm struggling here. > > > > I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I > > work for a University, where everyone has a quota. I'd literally have > > to create > 10K partitions. Is that really your intention? > > Yes. You'd group them all under a single filesystem in the hierarchy, > allowing you to manage NFS share options, compression, and more from a > single control point. > I'd agree except for backups. If the pools are going to grow beyond a reasonable-to-backup and reasonable-to-restore threshold (measured by the backup window), it would be practical to break it into smaller pools. After all, you'll probably have to restore a pool eventually. If that will take a week, your users won't be very happy with your solution. > > Of course, backups become a huge pain now. below, that's cumbersome > > for both backups and (especially) restores. > > Using traditional tools or ZFS send/receive? We are working on RFEs for > recursive snapshots, send, and recv, as well as preserving DSL > properties as part of a 'send', which should make backups of large > filesystem hierarchies much simpler. > Using EBS or NetBackup, can I get a single file back from tape only through the backup system? That's a big factor for production environments. Also, when users request a restore from tape from offsite backups, they'll usually specify a date range for when the file was 'good'. To accomplish that, you need to use the backup solution to find the requisite file. These 'fishing expeditions' (as I call them) can take a lot of time if direct access isn't available via the backup tool. ZFS basically eliminates the need to single file restores, because it has snapshots, then the user can have almost instant access to old copies of files. and is a lot quicker than even the fastest tape library. Just make daily snapshots and the need to restore a single file from tape is almost completely eliminated, you can still use netbackup for disasters, but to get access to a single old file a snapshot is much easier. You can also make it possible to have users initiate there own snapshots when they feel the need arises. James Dickens uadmin.blogspot.com I believe you're referring in the above to using zfs send/recv for backup to tape. Until the vendors work with zfs send/recv, it's not a viable option for filesystem backups in a production environment. Related to that, does anybody have a timeframe for direct support for ZFS send/recv (or something similar) in NBU or EBS? [ quota explanation deleted for brevity ] > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > ___ > zfs-discuss mailing list > zfs-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, May 18, 2006 at 12:46:28PM -0700, Charlie wrote: > Eric Schrock wrote: > > > Using traditional tools or ZFS send/receive? > > Traditional (amanda). I'm not seeing a way to dump zfs file systems to > tape without resorting to 'zfs send' being piped through gtar or > something. Even then, the only thing I could restore was an entire file > system. (We frequently restore single files for users...) Remember, ZFS is a fully POSIX-compliant filesystem. Any backup program that uses system calls to do its work will still function properly. Why would you believe that your backup program doesn't work with ZFS? Have you actually tried it? If it doesn't work, that's a big bug for us. > Perhaps, since zfs isn't limited to one snapshot per FS like fssnap is, > I should be redesigning everything. It sounds like I should look at > using many snapshots, and dumping to tape (each file system, somehow) > less frequently. That's definitely an option. You can also tell your backup program to not stop at filesystem boundaries so you can do entire trees of your namespace at once. --Bill ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Re: Re: zfs snapshot for backup, Quota
Bill Moore wrote: > On Thu, May 18, 2006 at 12:46:28PM -0700, Charlie wrote: >> Eric Schrock wrote: Using traditional tools or ZFS send/receive? >> Traditional (amanda). I'm not seeing a way to dump zfs file systems to >> tape without resorting to 'zfs send' being piped through gtar or >> something. Even then, the only thing I could restore was an entire file >> system. (We frequently restore single files for users...) > > Remember, ZFS is a fully POSIX-compliant filesystem. Any backup program > that uses system calls to do its work will still function properly. Why > would you believe that your backup program doesn't work with ZFS? Have > you actually tried it? If it doesn't work, that's a big bug for us. Of course, using system calls isn't an issue. Most backup systems funtion at a higher level than read() however :) I was thinking about amanda specifically, and I'd need zfsdump to do that. The result is thus: If I want incrementals, I must tell amanda to use tar. Using 'dump' is preferred for many reasons. And 'zfs send' is neat, but only mildly useful. -Charlie This message posted from opensolaris.org ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, 2006-05-18 at 16:43 -0500, James Dickens wrote: > On 5/18/06, Gregory Shaw <[EMAIL PROTECTED]> wrote: > > On Thu, 2006-05-18 at 12:12 -0700, Eric Schrock wrote: > > > On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: > > > > Sorry to revive such an old thread.. but I'm struggling here. > > > > > > > > I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I > > > > work for a University, where everyone has a quota. I'd literally have > > > > to create > 10K partitions. Is that really your intention? > > > > > > Yes. You'd group them all under a single filesystem in the hierarchy, > > > allowing you to manage NFS share options, compression, and more from a > > > single control point. > > > > > > > I'd agree except for backups. If the pools are going to grow beyond a > > reasonable-to-backup and reasonable-to-restore threshold (measured by > > the backup window), it would be practical to break it into smaller > > pools. > > > > After all, you'll probably have to restore a pool eventually. If that > > will take a week, your users won't be very happy with your solution. > > > > > > Of course, backups become a huge pain now. below, that's cumbersome > > > > for both backups and (especially) restores. > > > > > > Using traditional tools or ZFS send/receive? We are working on RFEs for > > > recursive snapshots, send, and recv, as well as preserving DSL > > > properties as part of a 'send', which should make backups of large > > > filesystem hierarchies much simpler. > > > > > > > Using EBS or NetBackup, can I get a single file back from tape only > > through the backup system? That's a big factor for production > > environments. Also, when users request a restore from tape from offsite > > backups, they'll usually specify a date range for when the file was > > 'good'. To accomplish that, you need to use the backup solution to find > > the requisite file. These 'fishing expeditions' (as I call them) can > > take a lot of time if direct access isn't available via the backup tool. > > ZFS basically eliminates the need to single file restores, because it > has snapshots, then the user can have almost instant access to old > copies of files. and is a lot quicker than even the fastest tape > library. Just make daily snapshots and the need to restore a single > file from tape is almost completely eliminated, you can still use > netbackup for disasters, but to get access to a single old file a > snapshot is much easier. > > You can also make it possible to have users initiate there own > snapshots when they feel the need arises. > > James Dickens > uadmin.blogspot.com > The above would be fine for testing. However, on an active filesystem that is more than 50% full, you'll find that large amounts of space will be used by the snapshots. We currently use a pair of the Bluearc Titan fileserver appliances. They have very similar snapshot functionality. Currently, we can't maintain more than about 3 days of snapshots every 4 hours due to space constraints. For filesystems that don't move much, 1 snapshot per day for a year may be practical. I doubt it, as snapshots have to be managed, and maintaining 365 snapshots per filesystem (not pool) will be very difficult. [ stuff deleted ] ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On the topic of ZFS snapshots: does the snapshot just capture the changed _blocks_, or does it effectively copy the entire file if any block has changed? That is, assuming that the snapshot (destination) stays inside the same pool space. -Erik ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Thu, May 18, 2006 at 03:41:13PM -0700, Erik Trimble wrote: > On the topic of ZFS snapshots: > > does the snapshot just capture the changed _blocks_, or does it > effectively copy the entire file if any block has changed? Incremental sends capture changed blocks. Snapshots capture all of the FS state as of the time the snapshot is taken, though it does so in constant time. Subsequent changes are kept as changed blocks, as deltas to the snapshot in the filesystem and clones. > That is, assuming that the snapshot (destination) stays inside the same > pool space. Of course it does. Er, what do you mean by 'destination'? ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
Just piqued my interest on this one - How would we enforce quotas of sorts in large filesystems that are shared? I can see times when I might want lots of users to use the same directory (and thus, same filesystem) but still want to limit the amount of space each user can consume. Thoughts? Nathan. :) On Fri, 2006-05-19 at 05:12, Eric Schrock wrote: > On Thu, May 18, 2006 at 11:42:58AM -0700, Charlie wrote: > > Sorry to revive such an old thread.. but I'm struggling here. > > > > I really want to use zfs. Fssnap, SVM, etc all have drawbacks. But I > > work for a University, where everyone has a quota. I'd literally have > > to create > 10K partitions. Is that really your intention? > > Yes. You'd group them all under a single filesystem in the hierarchy, > allowing you to manage NFS share options, compression, and more from a > single control point. > > > Of course, backups become a huge pain now. below, that's cumbersome > > for both backups and (especially) restores. > > Using traditional tools or ZFS send/receive? We are working on RFEs for > recursive snapshots, send, and recv, as well as preserving DSL > properties as part of a 'send', which should make backups of large > filesystem hierarchies much simpler. > > > Why can't we just have user quotas in zfs? :) > > The fact that per-user quotas exist is really a historical artifact. > With traditional filesystems, it is (effectively) impossible to have a > filesystem per user. The filesystem is a logical administrative control > point, allowing you to view usage, control properties, perform backups, > take snapshots etc. For home directory servers, you really want to do > these operations per-user, so logically you'd want to equate the two > (filesystem = user). Per-user quotas (the most common use of quotas, > but not the only one) were introduced because multiple users had to > share the same filesystem. > > ZFS quotas are intentionally not associated with a particular user > because a) it's the logical extension of "filesystems as control point", > b) it's vastly simpler to implement and, most importantly, c) separates > implementation from adminsitrative policy. ZFS quotas can be set on > filesystems which may represent projects, groups, or any other > abstraction, as well as on entire portions of the hierarchy. This allows > them to be combined in ways that traditional per-user quotas cannot. > > Hope that helps, > > - Eric > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > ___ > zfs-discuss mailing list > zfs-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss -- // // Nathan Kroenert [EMAIL PROTECTED] // // PTS EngineerPhone: +61 2 9844-5235 // // Sun ServicesDirect Ext: x57235 // // Level 2, 828 Pacific HwyFax:+61 2 9844-5311 // // Gordon2072 New South Wales Australia // // ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re[7]: [zfs-discuss] Re: Re: Due to 128KB limit in ZFS it can't saturate disks
Hello Roch, Monday, May 15, 2006, 3:23:14 PM, you wrote: RBPE> The question put forth is whether the ZFS 128K blocksize is sufficient RBPE> to saturate a regular disk. There is great body of evidence that shows RBPE> that the bigger the write sizes and matching large FS clustersize lead RBPE> to more throughput. The counter point is that ZFS schedules it's I/O RBPE> like nothing else seen before and manages to sature a single disk RBPE> using enough concurrent 128K I/O. Nevertheless I get much more throughput using UFS and writing with large block than using ZFS on the same disk. And the difference is actually quite big in favor of UFS. RBPE> at places. So I am proposing this for review by the community> RBPE> I first measured the throughput of a write(2) to raw device using for RBPE> instance this; RBPE> dd if=/dev/zero of=/dev/rdsk/c1t1d0s0 bs=8192k count=1024 RBPE> On Solaris we would see some overhead of reading the block from RBPE> /dev/zero and then issuing the write call. The tightest function that RBPE> fences the I/O is default_physio(). That function will issue the I/O to RBPE> the device then wait for it to complete. If we take the elapse time RBPE> spent in this function and count the bytes that are I/O-ed, this RBPE> should give a good hint as to the throughput thedevice is RBPE> providing. The above dd command will issue a single I/O at a time RBPE> (d-script to measure is attached). RBPE> Trying different blocksizes I see: RBPE>Bytes Elapse of phys IO Size RBPE>Sent RBPE> RBPE>8 MB; 3576 ms of phys; avg sz : 16 KB; throughput 2 MB/s RBPE>9 MB; 1861 ms of phys; avg sz : 32 KB; throughput 4 MB/s RBPE>31 MB; 3450 ms of phys; avg sz : 64 KB; throughput 8 MB/s RBPE>78 MB; 4932 ms of phys; avg sz : 128 KB; throughput 15 MB/s RBPE>124 MB; 4903 ms of phys; avg sz : 256 KB; throughput 25 MB/s RBPE>178 MB; 4868 ms of phys; avg sz : 512 KB; throughput 36 MB/s RBPE>226 MB; 4824 ms of phys; avg sz : 1024 KB; throughput 46 MB/s RBPE>226 MB; 4816 ms of phys; avg sz : 2048 KB; throughput 46 MB/s RBPE> 32 MB; 686 ms of phys; avg sz : 4096 KB; throughput 46 MB/s RBPE>224 MB; 4741 ms of phys; avg sz : 8192 KB; throughput 47 MB/s Just to be sure - you did reconfigure system to actually allow larger IO sizes? RBPE> Now lets see what ZFS gets. I measure using single dd process. ZFS RBPE> will chunk up data in 128K blocks. Now the dd command interact with RBPE> memory. But the I/O are scheduled under the control of spa_sync(). So RBPE> in the d-script (attached) I check for the start of an spa_sync and RBPE> time that based on elapse. At the same time I gather the number of RBPE> bytes and keep a count of I/O (bdev_strategy) that are being issued. RBPE> When the spa_sync completes we are sure that all those are on stable RBPE> storage. The script is a bit more complex because there are 2 threads RBPE> that issue spa_sync, but only one of them actuallybecomes RBPE> activated. So the script will print out some spurious lines of output RBPE> at times. I measure I/O with the script while this runs: RBPE> dd if=/dev/zero of=/zfs2/roch/f1 bs=1024k count=8000 RBPE> And I see: RBPE>1431 MB; 23723 ms of spa_sync; avg sz : 127 KB; throughput 60 MB/s RBPE>1387 MB; 23044 ms of spa_sync; avg sz : 127 KB; throughput 60 MB/s RBPE>2680 MB; 44209 ms of spa_sync; avg sz : 127 KB; throughput 60 MB/s RBPE>1359 MB; 24223 ms of spa_sync; avg sz : 127 KB; throughput 56 MB/s RBPE>1143 MB; 19183 ms of spa_sync; avg sz : 126 KB; throughput 59 MB/s RBPE> OK, I cheated. Here, ZFS is given a full disk to play with. In this RBPE> case ZFS enables the write cache. Note that even with the write cache RBPE> enabled, when the spa_sync() completes, it will be after a flush of RBPE> the cache has been executed. So the 60MB/sec do correspond to data set RBPE> to the platter. I just tried disabling the cache (with format -e) but RBPE> I am not sure if that is taken into account by ZFS; Results are the RBPE> same 60MB/sec. This will have to be confirmed. RBPE> With write cache enabled, the physio test reaches 66 MB/s as soon as RBPE> we are issuing 16KB I/Os. Here clearly though, data is not on the RBPE> platter when the timed function completes. RBPE> Another variable not fully controled is the physical (cylinder) RBPE> locations of the I/O. It could be that some of the differences come RBPE> from that. RBPE> What do I take away ? RBPE> a single 2MB physical I/O will get 46 MB/sec out of my disk. RBPE> 35 concurrent 128K I/O sustained followed by metadata I/O RBPE> followed by flush of the write cache allows ZFS to get 60 RBPE> MB/sec out of the same disk. RBPE> This is what underwrites my belief that 128K blocksize is sufficiently RBPE> large. Now, nothing here proves
Re: [zfs-discuss] Re: Re: zfs snapshot for backup, Quota
On Fri, 2006-05-19 at 10:18 +1000, Nathan Kroenert wrote: > Just piqued my interest on this one - > > How would we enforce quotas of sorts in large filesystems that are > shared? I can see times when I might want lots of users to use the same > directory (and thus, same filesystem) but still want to limit the amount > of space each user can consume. > > Thoughts? rats. > Nathan. :) OK :-) I've been wondering if I should mention this here, but I went ahead and blogged about it anyway. http://blogs.sun.com/roller/page/relling?entry=i_m_tired_of_owning Anyone who is really clever will easily get past a quota, especially at a university -- triple that probability for an engineering college. What it really boils down to is 2 things: 1. denial of service -- how to protect others from disk-hogs 2. contractual obligations -- how to charge the government (in the US anyway) for space used for government sponsored research... and pass the audit. A few years ago there was a 3rd thing: 3. how to pay for the disk space. Today, disk space is cheap. Really. All of the current college students I know carry around USB flash drives with all of their stuff on it. And iPods. If I were a college student, why would I risk my stuff being stored on the campus servers where "the man" might want to go snooping? Or, if you don't really care, use flickr, myspace, godaddy, gmail, or some other such storage service. Storage space really is becoming inexpensive. I'm not sure anybody can fix #2, but #1 can be accomplished within reason without resorting to user quotas. -- richard ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] ZFS recovery from a disk losing power
Since it's not exactly clear what you did with SVM I am assuming the following: You had a file system on top of the mirror and there was some I/O occurring to the mirror. The *only* time, SVM puts a device into maintenance is when we receive an EIO from the underlying device. So, in case a write occurred to the mirror, then the write to the powered off side failed (returned an EIO) and SVM kept going. Since all buffers sent to sd/ssd are marked with B_FAILFAST, the driver timeouts are low and the device is put into maintenance. If I understand Eric correctly, ZFS attempts to see if the device is really gone. However I am not quite sure what Eric means by: We currently only detect device failure when the device "goes away". Perhaps the issue here that ldi_open is successful when it should n't and therefore confusing ZFS. Another way to check is perform the same test, without any I/O occurring to the file system. Then run metastat -i (as root). This is similar to scrub for the volumes. -Sanjay Richard Elling wrote: On Tue, 2006-05-16 at 10:32 -0700, Eric Schrock wrote: On Wed, May 17, 2006 at 03:22:34AM +1000, grant beattie wrote: what I find interesting is that the SCSI errors were continuous for 10 minutes before I detached it, ZFS wasn't backing off at all. it was flooding the VGA console quicker than the console could print it all :) from what you said above, once per minute would have been more desirable. The "once per minute" is related to the frequency at which ZFS tries to reopen the device. Regardless, ZFS will try to issue I/O to the device whenever asked. If you believe the device is completely broken, the correct procedure (as documented in the ZFS Administration Guide), is to 'zpool offline' the device until you are able to repair it. I wonder why, given that ZFS knew there was a problem with this disk, that it wasn't marked FAULTED and the pool DEGRADED? This is the future enhancement that I described below. We need more sophisticated analysis than simply 'N errors = FAULTED', and that's what FMA provides. It will allow us to interact with larger fault management (such as correlating SCSI errors, identifying controller failure, and more). ZFS is a intentionally dumb. Each subsystem is responsible for reporting errors, but coordinated fault diagnosis has to happen at a higher level. [reason #8752, why pulling disk drives doesn't simulate real failures] There are also a number of cases where a successful or unsuccessful+retryable error codes carry the recommendation to replace the drive. There really isn't a clean way to write such diagnosis engines into the various file systems, LVMs, or databases which might use disk drives. Putting that intelligence into an FMA DE and tying that into file systems or LVMs is the best way to do this. -- richard ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] ZFS recovery from a disk losing power
On Thu, May 18, 2006 at 11:40:53PM -0600, Sanjay Nadkarni wrote: > Since it's not exactly clear what you did with SVM I am assuming the > following: > > You had a file system on top of the mirror and there was some I/O > occurring to the mirror. The *only* time, SVM puts a device into > maintenance is when we receive an EIO from the underlying device. So, > in case a write occurred to the mirror, then the write to the powered > off side failed (returned an EIO) and SVM kept going. Since all buffers > sent to sd/ssd are marked with B_FAILFAST, the driver timeouts are low > and the device is put into maintenance. the test was the same in both the SVM and the ZFS case. constant reads from the mirror device, and unplugging the power. the read throughput during this test with ZFS drops to around 20% until the device is manually removed from the pool, after which point it returns to normal. > If I understand Eric correctly, ZFS attempts to see if the device is > really gone. However I am not quite sure what Eric means by: > > >We currently only detect device failure when the device "goes away". > > Perhaps the issue here that ldi_open is successful when it should n't > and therefore confusing ZFS. yes, that seems to be the case. it appears to be caused by the way the aac card deals with the disk going away - it offlines the disk, and the LUN is still presented, but it now has zero length. also, after a disk is offlined by the card, there does not seem to be a way to tell the card to rescan the bus, so it requires a reboot (though there is nothing that ZFS can do which would fix that). I believe it can be done with the "aaccli" program provided by Adaptec, but that doesn't work with the Solaris-provided aac driver. > Another way to check is perform the same test, without any I/O > occurring to the file system. Then run metastat -i (as root). This is > similar to scrub for the volumes. with no IO activity on the mirror, metastat -i does not detect that anything is wrong. with IO activity, SVM offlines the metadevice when it gets a fatal error from the device. grant. ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss