On 03/08/2015 07:14 PM, Chris Murphy wrote:
On Sun, Mar 8, 2015 at 10:45 AM, Anton Aylward <opensuse@antonaylward.com> wrote:
On 03/08/2015 12:12 PM, don fisher wrote:
Why so many sub volumes.
You can choose not to have the ons you are given. You can choose to have other ones of your own making.
Are you sure? You've tested this thoroughly and you know in advance that snapper doesn't get confused or blow up? I'm pretty sure it has certain expectations what it's snapshotting and restoring.
Within the boundaries that I've discussed here in the past, yes i am sure. I have removed the subvolmes that came and replaced them with simple directories and mounted other file systems at that point. I have added a couple of subvolumes and I run zypper daily and it produces the snapshots. I've inspected them and they are what I expect. I tried a restore once and it worked. I've not used snappshotting for making backups. As i said earlier in this thread and elsewhere, I run zypper daily and it produces snappshots. Occasionally I flush them.
hey, this is Linux. Everything comes down to you choosing how to configure things.
That is not inherently a public good. It can be a significant barrier to entry and on-going maintenance. If the goal is to have a competitive desktop, it probably needs a default layout that's understandable, or who's going to maintain it?
The fact is that what gets 'maintained' is the capability to configure. Yes a default is there in the 'template' when you create a user account. But you can alter that template. The idea of being forced to have a specific panel with specific icons & widgets and not being able to customise is abhorrent. Not even Windows users would put up with that! You would not put up with the idea that all houses/home have to have the same rooms with the same furniture in the same position, would you? Away from the desktop things *have* to be confgurable. Apache, Postfix and more.
I'd be less concerned about nutty partition layouts or many Btrfs subvolumes, stateless installs were the norm. For example, my cyanogen phone has 28 partitions. This is not by my choice, this is the design of the phone, in fact I have no choice in the matter near as I can tell without likely breaking it. Does 28 GPT partitions sound nutty to me? Yes, quite. Doesn't it negatively impact me? No. If anything it probably helps separate things so they the thing is easy to reset, restore, and update.
I'm not sure there is an analogy here other than the phone is running a version of Linux. Cyanogen is a option you've taken, or it is for most phones; IIR there's only the 1+1 that has it as stock.
Very Soon Now, I'd like to see out of tree snapshots taken, and for that tree to be updated atomically, rather than the currently active fs tree. That way if the update blows up, just delete that whole tree. No rollback even necessary. Further, the update can happen in a chroot or container so its update environment is more stable/deterministic; and further my environment isn't the one doing the update or affected by it. So I don't have to reboot to get the offline update. It's an online update, and an optional reboot at my convenience to realize the benefit of that update.
I'd stop and think. Part of the reason the BtrFS people seem to want the whole of the system to be the single BtrFS tree is to do what you're talking about, since there is no other FS. Having the BtrFS deal with snapshotting a mounted ReiserFS or ext4FS leads to all manner of complications. As for the type of maintenance, well I've done that -- sort of -- and I've done even more back on AIX with the way updates and release and rollback was done there. The underlying model was more like LVM than BtrFS. I beleive there is work being done on a kernel that can be updated with without the need to reboot. How comprehensive updates there are I don't know: we will see.
Whether it is on your system is entirely up to you. You too can *choose* for it not to be so.
Choice without knowledge makes choice a coin toss. In order to make a meaningful choice requires knowledge.
I have no argument with statement. Sadly, we succumb to GUIs that try to embed such knowledge and really they are not 'expert systems' so much as the view of the programmer who put it together. Its one reason I prefer to use the command line for so many functions.
The more choices, the more knowledge. Acquiring knowledge takes time, and in that endeavor there is no short cut, no choice. So the expansion of choice over here inevitably leads to reduction of choice over there.
Perhaps, perhaps not. For example, I don't know all the CLI ways of manipulating KDE, the equivalent of using 'systemsettings'. But I do know where documentation on KDE lives, I do know where the config files live and I do know the basic patterns that that Linux uses, over and over. Yes there are various sets of patters; some config is in XML, some in well documented files like for Apache. You can get through better than 80% of it as a generalist. You just have to have the humility to RTFM or 'go google'. For the remaining 20% there are forums like this, boots and web sites. Yes, its a learning curve for those coming from Windows. Yes, OSX is pretty much for people who don't want to "look under the hood', though there are always people who will probe and hack anything - their car, their kitchen appliances, their bodies. The different distributions each have their edge, but my attitude is 'so what?' At various times in my life I've considered the 'baseline' UNIX to be a certain version and thought in terms of how other systems that I had to use differed from that baseline. UNIX/V7, BSD4.2, SUNOS, Solaris, AIX, fedora, Mandrake, and now Suse. I'm not denying that each vendor "adds value" in their own way, but I'm not obsessive and realise that the diversity is important. Its the same for house sizes and furnishing, cars, clothing, facial hair. Even in the military where everyone is wearing the same uniform there is incredible diversity.
e.g., all these knobs in Linux installers, inevitably lead to bugs.
You need to differentiate between bugs as programming errors and bugs at the operator level, people doing odd things in installation. As in 'that's a 4G drive so use GPT not MBR'. In my DatabaseofDotSigQuotes I have At least when humans go to casualty, they generally haven't gone into the Control Panel and messed with the settings... You point about knowledge is well taken, but there are really two points you miss. I've covered one, but want to amplify that learning required a degree of humility, and a lot of the people I encounter 'fiddling with the Control Panel' don't have that. The other is play and experimentation. Now before you talk about 'production systems' let me stop you. Production systems are not for playing with. I've had the unfortunate experience of seeing a manager hire an inexperienced sysadmin (for various reasons but mostly the guy 'came cheap') who pretty much managed to wipe out a running system, partly because there was no 'development' (aka non-critical) system to test the scripts for backup/restore that he was working on. I'd always taken it as axiomatic that you don't do development/testing on live machines, but neither the sysadmin nor the manager took that view. The sysadmin was fired. I think the manager was to blame. You could say it was about 'choice'. The manager chose to do things (by proxy) without understanding or bothering to learn. He was an arrogant sort would would not take advice, do basic risk management and put preventative or mitigating controls in place because he believed in the inerrant correctness and could not see why anyone would want to do things other than his way. “Imprimatur” and “ne varietur”.
Choice.
https://www.redhat.com/archives/fedora-devel-list/2008-January/msg00861.html The interesting this about that thread is that it walks a fine line between FOSS and the kind of formalisms about project management, scope definition and constraints, resource management and formal testing that one might associate with the traditional software development industry such as the old IBM, the old Oracle, the late 1990s Microsoft. No-one actually comes out and says that more formalism is needed, but it's there in the subtext. -- The Romans made their bridge-builders stand under their bridges. Is there a good reason why the software engineers of today don't have to entrust their lives to their code? -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org