ntfy_scripts/disks_checks.sh

75 lines
3.0 KiB
Bash
Raw Normal View History

2023-09-19 14:45:14 +02:00
#!/bin/bash
### I lost the link where I copied the first part of this script from, sorry, i'll update as soon as I find it back! The btrfs raid1 check part (second) and the btrfs device errors part (third) are written by tech_at_hangar.org
2023-09-19 14:45:14 +02:00
## The first part of this script checks for given minimum space on a mounted volume, and alerts ntfy in case the limit is surpassed.
## The second part of this script checks for missing disks on btrfs raid1 devices and alerts ntfy in case of missing devices
## The third part of this script checks for read/write errors on disks in mounted partitions and alerts ntfy in case of errors
2023-09-19 14:45:14 +02:00
# root volume
mingigs=10
avail=$(df | awk '$6 == "/" && $4 < '$mingigs' * 1024*1024 { print $4/1024/1024 }')
topicurl=https://NTFY_SERVER_IP/disk_alerts
if [ -n "$avail" ]; then
curl -k --retry 3 \
-d "Only $avail GB available on the root disk. Better clean that up." \
-H "Title: Low disk space alert on $(hostname)" \
-H "Priority: default" \
-H "Tags: warning,cd" \
$topicurl
fi
# OTHER VOLUME NAME
mingigs=500
avail=$(df | awk '$6 == "/mnt/OTHER_VOLUME" && $4 < '$mingigs' * 1024*1024 { print $4/1024/1024 }')
topicurl=https://NTFY_SERVER_IP/disk_alerts
if [ -n "$avail" ]; then
curl -k --retry 3 \
-d "Only $avail GB available on the /mnt/OTHER_VOLUME disk. Better clean that up." \
-H "Title: Low disk space alert on $(hostname)" \
-H "Priority: default" \
-H "Tags: warning,cd" \
$topicurl
fi
### check btrfs raid status on btrfs devices
2023-09-19 14:45:14 +02:00
if (btrfs fi show | grep -wic "Some devices missing"); then
echo Bad: - Btrfs is missing devices, some disk is broken
echo Sending alert...
curl -k --retry 3 \
-d "Bad - Some devices are missing in btrfs: maybe a broken disk?" \
-H "Title: Missing disks in btrfs on $(hostname)" \
-H "Priority: high" \
-H "Tags: warning,boar" \
$topicurl
fi
### check errors on btrfs devices
topicurl=https://NTFY_SERVER_IP/disk_alerts
# get list of btrfs mounted partition, put all results in a single line separated by a comma and remove the last comma
mounted_partitions="$(mount | grep btrfs | awk '{print $3}' | tr '\n' ',' | awk '{sub(/,$/,""); print}')"
for i in ${mounted_partitions//,/ }
do
# check btrfs errors on mounted device, if all output is 0 then return a single "1"
check_disk_errors=$(btrfs device stats $i | awk '{print $2}' | grep -c 0 -v)
if [ ${check_disk_errors} -eq 1 ] ; then
: # do nothing if output is a single "1"
else
echo Bad: - Btrfs is detecting errors on partition $i, some disk is about to be broken
echo Sending alert...
curl -k --retry 3 \
-d "Bad - Errors on btrfs mounted device $i: you will have to replace a disk. Check the status of mounted disks with: btrfs device stats $1" \
-H "Title: btrfs errors on partition $i of $(hostname)" \
-H "Priority: high" \
-H "Tags: warning,boar" \
$topicurl
fi
done