1 |
On 2019.01.07 14:35, Dale wrote: |
2 |
> Jack wrote: |
3 |
> > On 2019.01.07 05:46, Dale wrote: |
4 |
> >> Peter Humphrey wrote: |
5 |
> >> > On Sunday, 6 January 2019 22:13:31 GMT Dale wrote: |
6 |
> >> > |
7 |
> >> >> Even from my simple setup, LVM adds more benefits to managing |
8 |
> data |
9 |
> >> and |
10 |
> >> >> drives than it does risk. The biggest thing, placing blame |
11 |
> where it |
12 |
> >> >> lies. Blaming LVM for a drive dying is placing the blame on |
13 |
> >> something |
14 |
> >> >> that wasn't the root of the problem. The dying drive was the |
15 |
> >> problem, |
16 |
> >> >> using LVM or not. |
17 |
> >> > He isn't doing that, though. As I read it, he recounted the tale |
18 |
> of |
19 |
> >> recovering |
20 |
> >> > data from a failed drive, then imagined how much worse it would |
21 |
> be |
22 |
> >> if it were |
23 |
> >> > in an LVM setup. [Reported speech and mixed-up tenses causing me |
24 |
> a |
25 |
> >> problem |
26 |
> >> > here...] |
27 |
> >> > |
28 |
> >> > Thanks Gevisz, that was interesting. What we used to call a |
29 |
> >> cautionary tale. |
30 |
> >> > |
31 |
> >> |
32 |
> >> From what I've read, that can be overcome. If you get say a SMART |
33 |
> >> message that a drive is failing, just remove that drive or remove |
34 |
> the |
35 |
> >> whole LVM setup and use something else until a working drive setup |
36 |
> can |
37 |
> >> be made. Once ready, then move the data, if the drive still |
38 |
> works, to |
39 |
> >> the new drive. That is basically what I did when I swapped a |
40 |
> smaller |
41 |
> >> drive for a larger one. I moved the data from one drive to |
42 |
> another. It |
43 |
> >> did it fairly quickly. Someone posted that it may even be faster |
44 |
> to do |
45 |
> >> it with LVM's pvmove than it is with cp or rsync. I don't know |
46 |
> how true |
47 |
> >> that is but from what I've read, it moves the data really |
48 |
> efficiently. |
49 |
> >> If the drive has a very limited time before failure, speed is |
50 |
> >> important. If the drive is completely dead, replace the drive and |
51 |
> hope |
52 |
> >> the backups are good. Either way, LVM or not, a failing drive is a |
53 |
> >> failing drive. The data has to be moved if the drive still works |
54 |
> or the |
55 |
> >> data is gone if it just up and dies. The biggest thing, watching |
56 |
> the |
57 |
> >> SMART messages about the health of the drive. In the past when |
58 |
> I've had |
59 |
> >> a drive fail, I got error messages well ahead of time. On one |
60 |
> drive, I |
61 |
> >> removed the drive, set it aside, ordered a replacement drive, |
62 |
> installed |
63 |
> >> both drives and copied the data over. After I did all that, I |
64 |
> played |
65 |
> >> with the drive until it failed a day or so later. Lucky? Most |
66 |
> likely. |
67 |
> >> Still, it gave me time to transfer things over. |
68 |
> >> |
69 |
> >> While I get that LVM adds a layer to things, it also adds some |
70 |
> options |
71 |
> >> as well. Those options can prove helpful if one uses them. |
72 |
> >> |
73 |
> >> Just my thinking. |
74 |
> >> |
75 |
> >> Dale |
76 |
> > The only problem with all that is that SMART is far from completely |
77 |
> > reliable. I recently had a drive fail, and the resulting fsck on |
78 |
> the |
79 |
> > next reboot messed up many files. (Not a Gentoo system, although I |
80 |
> > don't think that made any difference.) After getting running |
81 |
> again, I |
82 |
> > did several SMART tests, including the full self-test, and it |
83 |
> reported |
84 |
> > ZERO errors. A few weeks later, it did the same thing, and shortly |
85 |
> > after that, it failed totally. I had done a few more full |
86 |
> self-tests |
87 |
> > before final failure, and all came back clean. I'd really love to |
88 |
> > find out there was something I did wrong in the testing, but I don't |
89 |
> > think so. I have not yet completely given up on trying to recover |
90 |
> > stuff from that drive, but as time goes on, there is less and less |
91 |
> > that I haven't rebuilt or replaced by re-downloading or changing |
92 |
> lost |
93 |
> > passwords, so it's less and less important. (That was a different |
94 |
> > drive from the one I messed up myself, as discussed in another |
95 |
> recent |
96 |
> > thread here.) |
97 |
> > |
98 |
> > Jack |
99 |
> > |
100 |
> |
101 |
> |
102 |
> But do you have any other way to get a warning? It may not work every |
103 |
> time, especially if the spindle motor just up and dies all of a sudden |
104 |
> but it does detect some errors. It is certainly better than having |
105 |
> nothing at all. So far, SMART has detected errors and warned me for |
106 |
> the |
107 |
> two drives I've had fail. My neighbor had a drive to fail and it gave |
108 |
> warnings as well, during boot up but SMART still spit our errors. |
109 |
> Thing |
110 |
> is, the owner ignored it until it wouldn't boot anymore. By that |
111 |
> time, |
112 |
> it was toast. They ran windoze. When SMART does warn, it pays to |
113 |
> listen. ;-) Mine emails me when any error is reported. |
114 |
> |
115 |
> Thing is, a bad drive will always risk the loss of data. Always has. |
116 |
> Monitoring SMART is better than nothing and generally gives some |
117 |
> warning. It's not perfect but there is nothing else that does any |
118 |
> better that I've heard or read about. It's the reason everyone should |
119 |
> back up data they can't afford to lose. |
120 |
> |
121 |
> Dale |
122 |
> |
123 |
> :-) :-) |
124 |
I do agree it is better than nothing, and I agree if SMART warns you, |
125 |
you better listen. I just wouldn't bet the farm (or even a small |
126 |
garden) on it. I'm coming closer and closer to just mirroring |
127 |
everything I can't easily recreate. It doubles my disk costs, but |
128 |
should save me some future grief. |