Slashdot Mirror
Gnome's Nautilus Gets ZFS Integration, In OpenSolaris
13bpower writes "Sun developer Erwann Chenede posted a new plugin for Nautilus that will integrate ZFS's backup capabilities with Nautilus. This should be a pretty killer feature." As one of the comments puts it, this adds a "Time Machine-esque" function to Solaris, through which a user can specify backup frequency, and when needed browse from available snapshots to restore files.
Why not go all the way?
Gnome's Nautilus: Gets "ZFS" Integration! (In OpenSolaris)
So what it gets integration with OpenSolaris.
You know what? Basic SMB mounting doesnt even work right in Ubuntu. Instead, it mounts crap via ~/.gvfs and fake handles for Gnome-only apps.
Instead, if Ubuntu used standard mounting techniques, it would mount as something sane like ~/mnt/$computer/$share_name/ so that all programs could use it easily.
Most likely, this ZFS setup uses the same non-standard techniques that make the features ONLY for gnome programs.
FAIL.
Some of us mangle "proper" punctuation and grammar all of the time because we write how we speak, we're not writing an English composition paper or a scientific journal. I find the headline to be as natural- sounding as possible while emphasizing that the implementation is an OpenSolaris-only one.
There's nothing wrong with having style. If you think this article's headline's is bad, then try to read Hunter S. Thompson or William S. Burroughs!
I don't use OpenSolaris, so I won't be seeing the benefits of this any time soon. (I once tried to install it, and couldn't work out the partitioning scheme, I don't install operating systems for fun any more.)
This is a great example of free software. Someone wants a feature, and they code it up.
Anyway, I was looking at getting something like this for MS Windows or Ubuntu the other month (instead of a proper version control system for people who couldn't cope). I found some interesting projects.
For example:
ext3cow, which you can use with a Time Traveling File Manager.
Copyfs is another versioned file system, and runs on FUSE.
In the end I didn't end up using either project, instead the usual folder with old version, as compared to folder with newer version is being used. (Curse tech illiteracy.)
Anyway, good work to those folks!
I wank in the shower.
That's great, except ZFS is currently stable only on Solaris 64 bit systems. The freebsd port is listed as alpha and is plain broken on 32 bit systems.
I, have, asthma, you, insensitive, clod.
Circumcision is child abuse.
The screen shot in the Sun blog shows Time Slider having 2.4 MB of snapshots available for possible recovery. I'm sure office workers writing memos and Slashdot posters whose posts regularly get eaten by the browser would love this feature. But what happens when the OS has to deal with the sort of big files churned out in a multimedia setting, say, a a multi-GB cache of digital video. Granted that someone dealing with such big files ought to make backups using other means, the question remains: does ZFS have the intelligence, for example, to preserve smaller files when one or two big files are threatening to exhaust filesystem's allotted snapshot space? Which file gets priority?
let's be clear. poor grammar is irritating. snarking about poor grammar is also irritating.
with that in mind, mistaking sloppiness with or ignorance of standard grammar for "style" certainly reflects odd if a lack of taste.
Thompson, Burroughs, Joyce (take your pick) didn't stylistically torture grammar because they were sloppy or unaware of accepted practices.
The difference is similar to the difference between Faulkner's prose and the sometimes painful and sometimes hilarious parodies of his prose in that famous contest.
it's all good right? right? given that both style and taste are kind of snooty subjective affectations. i am snarking but goodheartedly. i think. ;)
... some time ago for a usable time-machine substitute for linux, but failed miserably on the gnome integration.
;).
;), but it's python anyway. Right now the project is somewhat unmantained though, by the mentioned lack of time and some tiredness from the project.
Unfortunately when we tried we were informed by the kind folk of the nautilus mailing list that it would be pretty difficult, as it's not feasible by using the python nautilus-extensions, and we would have to rewrite a lot of code. Gvfs was not ready yet (and mostly undocumented) so our code would be instantly deprecated, and many of the supposed options of a quick google were hopeleslly outdated / broken / unsupported / deprecated -such as the bonobo views, which I couldn't get to work at all.
Finally my personal struggles with autotools (devilish tool!) ended up burying the idea. Pity, but I'm glad that the opensolaris folk were able to do it. Their GUI rocks a lot IMHO (I tried to congratulate them on their blog but the post got lost somehow
Perhaps if one of us gets enough free time in RL we'll port their patches (should be GPL I guess) to our linux implementation, in a cleaner nautilus C plugin.
BTW feel free to try the application, it should be functional (as in "won't break your hard drive down nor delete anything it should, though it might take some space in ~/.hdlorean"), though not as polished as we would have liked. We even managed to build a debian package
I know it's not zfs-related, but nautilus is such a freaking *hog* that I can't avoid this opportunity to *beg* nautilus developers to lighten their memory footprint.
I've been able to do better than time-machine feature-wise since snapshotting was online with zfs.
I've had automated jobs taking snapshots every 5 minutes and keeping them for as long as I had space available for them (sitting at around 9 months worth atm).
The script rotates through the snapshots, and passes through the snapshots where no changes were made and marking them for my approval to remove, just to try and keep the listings tolerable...(12 * 24 * 275)....
From there I wrote a routine to allow me to easily (though not always speedily) look for the file that I use as a parameter and give me a listing of all the revisions available, which I can select to *restore* or copy back into place under the same name, a different name, etc...
I can do this with entire directory trees, and am working on routines to detect if trees were *moved* via SMB shares or other misadventure (or intentionally) and hopefully be able to then offer up the option to move either the current iteration of the moved directory back, or an old version.
It's been fun writing the utilities, and yes, they are all command line, and all using my old favorite scripting language, ksh.
Who is general failure, and why is he reading my hard drive?
That does sound quite cool, but it doesn't really solve the problem for your average office worker who just wants to get that file they deleted back.
Providing a GUI for it frees up IT support to do something other then run a bunch of command line scripts to find the older version, and also makes it possible for users to find out that it's even possible to do this.
The statements of the parent poster are not correct.
If a user requests an unmounted drive through a GIO-aware application, all mounting and unmounting operations are done by GIO. If necessary, the user is presented with a login-dialog.
For convenience all active GIO drives are mapped onto the filesystem through FUSE. Even GIO unaware drives can access remote drives this way, though they are limited to POSIX file system calls.
Handles for GIO drives do all relate to the file system mapping, so when you request a file handle from GIO, it will give you a path reference that will work with all applications, not just GLib ones.
GIO will automatically bypass the FUSE mapping if a GIO handle is requested from a path reference. This has been implemented in the last few month.
Mounting the user mounts in /mnt does not make much sense in multi-user environments. You don't necessary want other users to allow to access remote mounts with your user credentials.
Mounting remove file systems permanently is a system administrator task, not a user task.
The only valid complaint could be formulated as:
"Ubuntu doesn't provide a convenient way to set up permanent remote mounts."
Instead the parent poster, failing to identify the problem - (there is no standard unix way to handle the kind of thing ZFS does), chooses to attack SUN's engineers for using GNOME technologies to implement a prototype.
I have no clue why this post got modded "+5, Interesting".
Hmmm... Yes a gui would be nice, and someday there may be one... but for now, on my personal Solaris system at home, I don't need one - although I'd like to get the RAM up to 64GB on that thing for a better all around ZFS experience...
It's kind of nice booting up multiple xVM virtual PCs running various operating systems and barely using the disk at all...
Windows XP (after initial load, and running, can reboot and operate with almost zero disk I/O running off the ARC) - Linux is almost none as well...
It makes a very nice virtual workstation server...
Who is general failure, and why is he reading my hard drive?
What sort of hardware are you running on? Having seen some of the features that Solaris provides (especially ZFS, and Zones), I'd love to set up a server to play with it on, but I have a feeling I'm going to need some sort of exotic hardware to get started.
Well, I happened to have fallen into an Intel test platform that sat around for a year or two..
Supermicro server chassis, two dual core Xeon 64-bit processors (2.x ghz), 12 GB DDR3 RAM (with enough slots left that I can take it to 16GB with 1GB sticks), dual adaptec ultra 320 controllers, 6 sata controllers (on the motherboard), currently configured with 8 x 75GB drives - with the intent to replace them with terabyte (900ish GiB) sata drives.
I need to get a 2nd 700watt hot-swap psu and another 1400VA ups to drive the 2nd supply.
I love the system, I do not love the electric bill... =)
At some point, I'll retire it in favor of an old gaming config with external sata drives or something like that...
I've also had success running Solaris (open and standard) on IBM/Lenovo thinkpad laptops, as well as AMD hardware of the homebuilt and HP/Compaq variety.
Still issues (last time I tried was with Solaris 10, U5) with IBM RaidServ (IBM hijacked Adaptec raid controllers) controllers.
Who is general failure, and why is he reading my hard drive?