Greetings
This is the first time in my life I'm writing on a linux mailing list,
hope I'm doing things right :)
I'm setting up a NAS system on a raspberry PI, I want the data to be
stored on a RAID-5 setup. I currently have 3 2TB WD-red drives
assembled into a RAID-5, connected to the PI via a USB-SATA adapter
and an external power source (which also powers the PI by the way).
The issue is this: sometimes it happens, that when you move the PI or
just slightly touch the cabling, that a drive would disappear from the
system, to suddenly reappear, as if it was plugged out and immediatly
plugged in again. This of course registeres in mdadm as a failed
drive, even though all data is still correct, and it could be
reassembled to a fully functioning array (event counts are the same in
the superblock), in fact running `mdam --assemble --force /dev/md127
<devices>` puts the array back in its normal running state (If event
counts still match on all the disks).
Only problem arises if this was to happen during a write to the array,
in which case the array will probably have to be rebuilt, which takes
2-3 days because of the PI-s bad IO capabilyties. This is the scenario
I want to avoid whenever possible.
I would like to set up mdadm in a way that it automatically puts an
array into read-only mode when a drive is failing, or being forcefully
removed, or when a drive is missing, to maximaize the chances of a
successfull reassembly without a rebuild (since the data is still
there and in tact, the drives arent dead, its just a bad cabling,which
I cant fix to a 100% right now)
I've written a script, and set mdadm up to call it with --monitor:
```
#!/bin/bash
logfile="/home/ganter/temp/mdadmEvents.log"
echo ".--==== $(date) ====--." >> "$logfile"
echo "$@" >> "$logfile"
case "$1" in
"Fail")
echo "sudo mdadm --readonly $2" 2>&1 >> "$logfile"
mdadm --readonly "$2"
;;
"DegradedArray")
echo "mdadm --readonly $2" 2>&1 >> "$logfile"
mdadm --readonly "$2"
;;
esac
```
but its as if the call `mdadm --readonly` gets ignored, everything
else works as expected (the switch, the echo calls), showing up in the
logfile, its just that the array stays in readwrite mode, which in my
case is very very bad.
Logging `echo "$(which mdadm)" >> "$logfile"` shows fine
(/usr/bin/mdadm), so the command is found, also `echo "$(mdadm -D
/dev/md127)" >> "$logfile"` works the same as if run from terminal,
echo `"$(whoami)" >> "$logfile"` says I'm root. I really don't
understand what could be wrong here, is mdadm set up to somehow filter
and ignore calling itself like this from the monitoring program for
some reason? Or am I missing something.
Running the script manually from the command line passing arguments
like so: `script.sh Fail /dev/md127` works just fine, the array
switches to read only mode.
No amount of google-ing helped me in this regard, I didn't find any
questions regarding this topic anywhere, this mailing list is my last
hope. Any advice is appreciated :)
I'm currently not running the system on the PI, all of the above
happened on an Arch Linux distro with KDE desktop, if that counts in
any way.
Best regards, thank you for your time
