1 |
On Sunday, 17 March 2019 17:15:26 GMT Mick wrote: |
2 |
> On Sunday, 17 March 2019 17:05:58 GMT Grant Taylor wrote: |
3 |
> > On 3/17/19 10:48 AM, Peter Humphrey wrote: |
4 |
> > > My little Atom box has a small rescue system which I boot once a week |
5 |
> > > to back up the main system. The backup script is a simple list of bash |
6 |
> > > commands to mount partitions and tar them to a USB disk. |
7 |
> > |
8 |
> > Please share a copy of the backup script. |
9 |
|
10 |
I've found the problem. I wrote the script many years ago, before cgroups were |
11 |
invented, and more recently my script was unmounting /sys/fs/cgroup/ along |
12 |
with the real partitions. I can't begin to imagine what chaos that would |
13 |
cause, but having squashed the bug I don't see the self-unmounting any more. |
14 |
|
15 |
The script starts with everything but the root filesystem unmounted, then |
16 |
mounts one partition at a time to tar it to a USB disk. The problem was a |
17 |
side-effect of the way I was first unmounting everything (don't ask). |
18 |
|
19 |
Sorry about the noise. It did cause me to look more carefully though. |
20 |
|
21 |
> > > While the backup is running I run another script to clean up after |
22 |
> > > any recent update, which involves removing surplus packages, running |
23 |
> > > eclean etc. |
24 |
> > |
25 |
> > Please share a copy of your cleanup script. |
26 |
> > |
27 |
> > > But! At least once per session I have to remount the root filesystem |
28 |
> > > read- |
29 |
> > > write because something, presumably tar, has caused it to be remounted |
30 |
> > > read- only. |
31 |
> > |
32 |
> > "tar" itself shouldn't alter mounts at all. |
33 |
> > |
34 |
> > It is possible that there could be a mount option that causes the file |
35 |
> > system to be remounted read-only if there is a problem accessing the |
36 |
> > file system. |
37 |
> > |
38 |
> > > Where do I start tracking this down? This behaviour was a factor in |
39 |
> > > my suspecting an SSD failure, but I've replaced that and still get the |
40 |
> > > same remounting. |
41 |
> > |
42 |
> > But such remounting is not likely on a new SSD. |
43 |
> |
44 |
> Yes, it would be unlikely. |
45 |
> |
46 |
> > I'd need to see the scripts to even hazard a guess as to what might be |
47 |
> > happening. |
48 |
> |
49 |
> Also, in your previous thread you mentioned you were about to run memtest to |
50 |
> discard the possibility of a faulty RAM. Did you run it overnight and what |
51 |
> did you get? |
52 |
|
53 |
After fixing my backup script I did run memtest through a few cycles, but it |
54 |
didn't find anything wrong. |
55 |
|
56 |
-- |
57 |
Regards, |
58 |
Peter. |