1. #!/bin/bash
2.  
3. ###### ZPool & SMART status report with FreeNAS config backup
4. ### Original script by joeschmuck, modified by Bidelu0hm, then by melp (me)
5.  
6. ### At a minimum, enter email address in user-definable parameter section. Feel free to edit other user parameters as needed.
7. ### If you find any errors, feel free to contact me on the FreeNAS forums (username melp) or email me at jason at jro dot io.
8.  
9. ### Version: v1.3
10. ### Changelog:
11. # v1.3:
12. #   - Added scrub duration column
13. #   - Fixed for FreeNAS 11.1 (thanks reven!)
14. #   - Fixed fields parsed out of zpool status
15. #   - Buffered zpool status to reduce calls to script
16. # v1.2:
17. #   - Added switch for power-on time format
18. #   - Slimmed down table columns
19. #   - Fixed some shellcheck errors & other misc stuff
20. #   - Added .tar.gz to backup file attached to email
21. #   - (Still coming) Better SSD SMART support
22. # v1.1:
23. #   - Config backup now attached to report email
24. #   - Added option to turn off config backup
25. #   - Added option to save backup configs in a specified directory
26. #   - Power-on hours in SMART summary table now listed as YY-MM-DD-HH
27. #   - Changed filename of config backup to exclude timestamp (just uses datestamp now)
28. #   - Config backup and checksum files now zipped (was just .tar before; now .tar.gz)
29. #   - Fixed degrees symbol in SMART table (rendered weird for a lot of people); replaced with a *
30. #   - Added switch to enable or disable SSDs in SMART table (SSD reporting still needs work)
31. #   - Added most recent Extended & Short SMART tests in drive details section (only listed one before, whichever was more recent)
32. #   - Reformatted user-definable parameters section
33. #   - Added more general comments to code
34. # v1.0:
35. #   - Initial release
36.  
37. ### TODO:
38. # - Fix SSD SMART reporting
39.                                                                                
40. # - Add support for conveyance test
41.  
42.  
43. ###### User-definable Parameters
44. ### Email Address
45. email="morten@strandbo.org"
46.  
47. ### Global table colors
48. okColor="#c9ffcc"       # Hex code for color to use in SMART Status column if drives pass (default is light green, #c9ffcc)
49. warnColor="#ffd6d6"     # Hex code for WARN color (default is light red, #ffd6d6)
50. critColor="#ff0000"     # Hex code for CRITICAL color (default is bright red, #ff0000)
51. altColor="#f4f4f4"      # Table background alternates row colors between white and this color (default is light gray, #f4f4f4)
52.  
53.  
54. ### SMART status summary table settings
55. includeSSD="true"      # [NOTE: Currently this is pretty much useless] Change to "true" to include SSDs in SMART status summary table; "false" to disable
56. tempWarn=40             # Drive temp (in C) at which WARNING color will be used
57. tempCrit=45             # Drive temp (in C) at which CRITICAL color will be used
58. sectorsCrit=10          # Number of sectors per drive with errors before CRITICAL color will be used
59. testAgeWarn=7           # Maximum age (in days) of last SMART test before CRITICAL color will be used
60. powerTimeFormat="ymdh"  # Format for power-on hours string, valid options are "ymdh", "ymd", "ym", or "y" (year month day hour)
61.  
62.  
63. ###### Auto-generated Parameters
64. host=$(hostname -s)
65. logfile="/tmp/smart_report.tmp"
66. subject="Status Report for ${host}"
67. boundary="gc0p4Jq0M2Yt08jU534c0p"
68. if [ "$includeSSD" == "true" ]; then
69.     drives=$(for drive in $(/bin/lsblk -d | awk  '{ print $1 }' | awk '{ if(NR>1)print }' | sort); do
70.         if [ "$(/usr/sbin/smartctl -i /dev/"${drive}" | grep "SMART support is: Enabled")" ]; then
71.             printf "%s " "${drive}"
72.         fi
73.     done | awk '{for (i=NF; i!=0 ; i--) print $i }')
74. else
75.     drives=$(for drive in $(/bin/lsblk -d | awk  '{ print $1 }' | awk '{ if(NR>1)print }' | sort); do
76.         if [ "$(/usr/sbin/smartctl -i /dev/"${drive}" | grep "SMART support is: Enabled")" ] && ! [ "$(/usr/sbin/smartctl -i /dev/"${drive}" | grep "Solid State Device")" ]; then
77.             printf "%s " "${drive}"
78.         fi
79.     done | awk '{for (i=NF; i!=0 ; i--) print $i }')
80. fi
81.  
82.  
83.  
84. ###### Email pre-formatting
85. ### Set email headers
86. (
87.     echo "From: ${email}"
88.     echo "To: ${email}"
89.     echo "Subject: ${subject}"
90.     echo "MIME-Version: 1.0"
91.     echo "Content-Type: multipart/mixed; boundary=${boundary}"
92. ) > "$logfile"
93.  
94. (
95.  echo "--${boundary}"
96.  echo "Content-Type: text/html"
97. ) >> "$logfile"
98.  
99. ###### Report Summary Section (html tables)
100. ### SMART status summary table
101. (
102.     # Write HTML table headers to log file
103.     echo "<br><br>"
104.     echo "<table style=\"border: 1px solid black; border-collapse: collapse;\">"
105.     echo "<tr><th colspan=\"15\" style=\"text-align:center; font-size:20px; height:40px; font-family:courier;\">SMART Status Report Summary</th></tr>"
106.     echo "<tr>"
107.     echo "  <th style=\"text-align:center; width:100px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Device</th>"
108.     echo "  <th style=\"text-align:center; width:130px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Serial<br>Number</th>"
109.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">SMART<br>Status</th>"
110.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Temp</th>"
111.     echo "  <th style=\"text-align:center; width:120px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Power-On<br>Time</th>"
112.     echo "  <th style=\"text-align:center; width:100px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Start/Stop<br>Count</th>"
113.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Spin<br>Retry<br>Count</th>"
114.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Realloc'd<br>Sectors</th>"
115.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Realloc<br>Events</th>"
116.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Current<br>Pending<br>Sectors</th>"
117.     echo "  <th style=\"text-align:center; width:120px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Offline<br>Uncorrectable<br>Sectors</th>"
118.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">UltraDMA<br>CRC<br>Errors</th>"
119.     echo "  <th style=\"text-align:center; width:80px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Seek<br>Error<br>Health</th>"
120.     echo "  <th style=\"text-align:center; width:100px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Last Test<br>Age (days)</th>"
121.     echo "  <th style=\"text-align:center; width:100px; height:60px; border:1px solid black; border-collapse:collapse; font-family:courier;\">Last Test<br>Type</th></tr>"
122.     echo "</tr>"
123. ) >> "$logfile"
124. for drive in $drives; do
125.     (
126.         # For each drive detected, run "smartctl -A -i" and parse its output. This whole section is a single, long statement, so I'll make all comments here.
127.         # Start by passing awk variables (all the -v's) used in other parts of the script. Other variables are calculated in-line with other smartctl calls.
128.         # Next, pull values out of the original "smartctl -A -i" statement by searching for the text between the //'s.
129.         # After parsing the output, compute other values (last test's age, on time in YY-MM-DD-HH).
130.         # After these computations, determine the row's background color (alternating as above, subbing in other colors from the palate as needed).
131.         # Finally, print the HTML code for the current row of the table with all the gathered data.
132.         /usr/sbin/smartctl -A -i /dev/"$drive" | \
133.         awk -v device="$drive" -v tempWarn="$tempWarn" -v tempCrit="$tempCrit" -v sectorsCrit="$sectorsCrit" -v testAgeWarn="$testAgeWarn" \
134.         -v okColor="$okColor" -v warnColor="$warnColor" -v critColor="$critColor" -v altColor="$altColor" -v powerTimeFormat="$powerTimeFormat" \
135.         -v lastTestHours="$(/usr/sbin/smartctl -l selftest /dev/"$drive" | grep "# 1" | awk '{print $9}')" \
136.         -v lastTestType="$(/usr/sbin/smartctl -l selftest /dev/"$drive" | grep "# 1" | awk '{print $3}')" \
137.         -v smartStatus="$(/usr/sbin/smartctl -H /dev/"$drive" | grep "SMART overall-health" | awk '{print $6}')" ' \
138.         /Serial Number:/{serial=$3} \
139.         /Temperature_Celsius/{temp=($10 + 0)} \
140.         /Power_On_Hours/{onHours=$10} \
141.         /Start_Stop_Count/{startStop=$10} \
142.         /Spin_Retry_Count/{spinRetry=$10} \
143.         /Reallocated_Sector/{reAlloc=$10} \
144.         /Reallocated_Event_Count/{reAllocEvent=$10} \
145.         /Current_Pending_Sector/{pending=$10} \
146.         /Offline_Uncorrectable/{offlineUnc=$10} \
147.         /UDMA_CRC_Error_Count/{crcErrors=$10} \
148.         /Seek_Error_Rate/{seekErrorHealth=$4} \
149.         END {
150.             testAge=int((onHours - lastTestHours) / 24);
151.             yrs=int(onHours / 8760);
152.             mos=int((onHours % 8760) / 730);
153.             dys=int(((onHours % 8760) % 730) / 24);
154.             hrs=((onHours % 8760) % 730) % 24;
155.             if (powerTimeFormat == "ymdh") onTime=yrs "y " mos "m " dys "d " hrs "h";
156.             else if (powerTimeFormat == "ymd") onTime=yrs "y " mos "m " dys "d";
157.             else if (powerTimeFormat == "ym") onTime=yrs "y " mos "m";
158.             else if (powerTimeFormat == "y") onTime=yrs "y";
159.             else onTime=yrs "y " mos "m " dys "d " hrs "h ";
160.             if ((substr(device,3) + 0) % 2 == 1) bgColor = "#ffffff"; else bgColor = altColor;
161.             if (smartStatus != "PASSED") smartStatusColor = critColor; else smartStatusColor = okColor;
162.             if (temp >= tempCrit) tempColor = critColor; else if (temp >= tempWarn) tempColor = warnColor; else tempColor = bgColor;
163.             if (spinRetry != "0") spinRetryColor = warnColor; else spinRetryColor = bgColor;
164.             if ((reAlloc + 0) > sectorsCrit) reAllocColor = critColor; else if (reAlloc != 0) reAllocColor = warnColor; else reAllocColor = bgColor;
165.             if (reAllocEvent != "0") reAllocEventColor = warnColor; else reAllocEventColor = bgColor;
166.             if ((pending + 0) > sectorsCrit) pendingColor = critColor; else if (pending != 0) pendingColor = warnColor; else pendingColor = bgColor;
167.             if ((offlineUnc + 0) > sectorsCrit) offlineUncColor = critColor; else if (offlineUnc != 0) offlineUncColor = warnColor; else offlineUncColor = bgColor;
168.             if (crcErrors != "0") crcErrorsColor = warnColor; else crcErrorsColor = bgColor;
169.             if ((seekErrorHealth + 0) < 100) seekErrorHealthColor = warnColor; else seekErrorHealthColor = bgColor;
170.             if (testAge > testAgeWarn) testAgeColor = warnColor; else testAgeColor = bgColor;
171.             printf "<tr style=\"background-color:%s;\">\n" \
172.                 "<td style=\"text-align:center; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">/dev/%s</td>\n" \
173.                 "<td style=\"text-align:center; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
174.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
175.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%d*C</td>\n" \
176.                 "<td style=\"text-align:center; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
177.                 "<td style=\"text-align:center; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
178.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
179.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
180.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
181.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
182.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
183.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
184.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s%%</td>\n" \
185.                 "<td style=\"text-align:center; background-color:%s; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%d</td>\n" \
186.                 "<td style=\"text-align:center; height:25px; border:1px solid black; border-collapse:collapse; font-family:courier;\">%s</td>\n" \
187.             "</tr>\n", bgColor, device, serial, smartStatusColor, smartStatus, tempColor, temp, onTime, startStop, spinRetryColor, spinRetry, reAllocColor, reAlloc, \
188.             reAllocEventColor, reAllocEvent, pendingColor, pending, offlineUncColor, offlineUnc, crcErrorsColor, crcErrors, seekErrorHealthColor, seekErrorHealth, \
189.             testAgeColor, testAge, lastTestType;
190.         }'
191.     ) >> "$logfile"
192. done
193. # End SMART summary table and summary section
194. (
195.     echo "</table>"
196.     echo "<br><br>"
197. ) >> "$logfile"
198.  
199. # Drive usage
200. (
201.         echo "<b>RAID Usage</b><br>"
202.         echo "<pre>"
203.         df -h /dev/md0
204.         echo "</pre>"
205.         echo "<br>"
206. ) >> "$logfile"
207.  
208.  
209. ###### Detailed Report Section (monospace text)
210. echo "<pre style=\"font-size:14px\">" >> "$logfile"
211.  
212. ### SMART status for each drive
213. for drive in $drives; do
214.     # Gather brand and serial number of each drive
215.     brand="$(/usr/sbin/smartctl -i /dev/"$drive" | grep "Model Family" | awk '{print $3, $4, $5}')"
216.     serial="$(/usr/sbin/smartctl -i /dev/"$drive" | grep "Serial Number" | awk '{print $3}')"
217.     (
218.         # Create a simple header and drop the output of some basic smartctl commands
219.         echo "<br>"
220.         echo "<b>########## SMART status report for ${drive} drive (${brand}: ${serial}) ##########</b>"
221.         /usr/sbin/smartctl -H -A -l error /dev/"$drive"
222.         /usr/sbin/smartctl -l selftest /dev/"$drive" | grep "Extended \\|Num" | cut -c6- | head -2
223.         /usr/sbin/smartctl -l selftest /dev/"$drive" | grep "Short \\|Num" | cut -c6- | head -2 | tail -n -1
224.         echo "<br><br>"
225.     ) >> "$logfile"
226. done
227. # Remove some un-needed junk from the output
228. sed -i '' -e '/smartctl 6.6/d' "$logfile"
229. sed -i '' -e '/Copyright/d' "$logfile"
230. sed -i '' -e '/=== START OF READ/d' "$logfile"
231. sed -i '' -e '/SMART Attributes Data/d' "$logfile"
232. sed -i '' -e '/Vendor Specific SMART/d' "$logfile"
233. sed -i '' -e '/SMART Error Log Version/d' "$logfile"
234.  
235. ### End details section, close MIME section
236. (
237.     echo "</pre>"
238.     echo "--${boundary}--"
239. )  >> "$logfile"
240.  
241. ### Send report
242. /usr/sbin/sendmail -t -oi < "$logfile"
243. #rm "$logfile"
244.  
245.  
246.