1
0
mirror of https://github.com/Spearfoot/FreeNAS-scripts synced 2025-12-10 05:13:53 +00:00

Enhancements for new SAS support

Added counts for the SATA and SAS drive lists and modified the code to gracefully emit either or both types in the report.
This commit is contained in:
Keith Nash
2020-07-03 12:39:25 -05:00
committed by GitHub
parent 8ba42e8828
commit 335085177f

View File

@@ -10,7 +10,7 @@ smartctl=/usr/local/sbin/smartctl
freenashost=$(hostname -s | tr '[:lower:]' '[:upper:]') freenashost=$(hostname -s | tr '[:lower:]' '[:upper:]')
boundary="===== MIME boundary; FreeNAS server ${freenashost} =====" boundary="===== MIME boundary; FreeNAS server ${freenashost} ====="
logfile="/tmp/smart_report.tmp" logfile="smart_report.tmp"
subject="SMART Status Report for ${freenashost}" subject="SMART Status Report for ${freenashost}"
tempWarn=40 tempWarn=40
tempCrit=45 tempCrit=45
@@ -18,75 +18,54 @@ sectorsCrit=10
testAgeWarn=1 testAgeWarn=1
warnSymbol="?" warnSymbol="?"
critSymbol="!" critSymbol="!"
Drive_count=0
SATA_count=0
SAS_count=0
Drive_list=""
SATA_list=""
SAS_list=""
# We need a list of the SMART-enabled drives on the system. Choose one of these # Get list of SMART-enabled drives
# three methods to provide the list. Comment out the two unused sections of code.
# 1. A string constant; just key in the devices you want to report on here:
#drives="/dev/da1 /dev/da2 /dev/da3 /dev/da4 /dev/da5 /dev/da6 /dev/da7 /dev/da8 /dev/ada0"
# 2. A systcl-based technique suggested on the FreeNAS forum:
#drives=$(for drive in $(sysctl -n kern.disks); do \
#if [ "$("${"$smartctl"}" -i /dev/${drive} | grep "SMART support is: Enabled" | awk '{print $3}')" ]
#then printf ${drive}" "; fi done | awk '{for (i=NF; i!=0 ; i--) print $i }')
# 3. "$smartctl"-based functions:
get_smart_drives() get_smart_drives()
{ {
gs_smartdrives=""
gs_drives=$("$smartctl" --scan | awk '{print $1}') gs_drives=$("$smartctl" --scan | awk '{print $1}')
for gs_drive in $gs_drives; do for gs_drive in $gs_drives; do
gs_smart_flag=$("$smartctl" -i "$gs_drive" | grep -E "SMART support is:[[:blank:]]+Enabled" | awk '{print $4}') gs_smart_flag=$("$smartctl" -i "$gs_drive" | grep -E "SMART support is:[[:blank:]]+Enabled" | awk '{print $4}')
if [ "$gs_smart_flag" = "Enabled" ]; then if [ "$gs_smart_flag" = "Enabled" ]; then
gs_smartdrives="$gs_smartdrives $gs_drive" Drive_list="$Drive_list $gs_drive"
Drive_count=$((Drive_count + 1))
fi fi
done done
echo "$gs_smartdrives"
} }
drives=$(get_smart_drives) # Get list of SATA disks
# Checks whether it is a SATA disk
get_sata_drives() get_sata_drives()
{ {
gsata_smartdrives="" for drive in $Drive_list; do
for drive in $drives; do
gsata_smart_flag=$("$smartctl" -i "$drive" | grep -E "SATA Version is:[[:blank:]]" | awk '{print $4}') gsata_smart_flag=$("$smartctl" -i "$drive" | grep -E "SATA Version is:[[:blank:]]" | awk '{print $4}')
if [ "$gsata_smart_flag" = "SATA" ]; then if [ "$gsata_smart_flag" = "SATA" ]; then
gsata_smartdrives="$gsata_smartdrives $drive" SATA_list="$SATA_list $drive"
SATA_count=$((SATA_count + 1))
fi fi
done done
echo "$gsata_smartdrives"
} }
satadrives=$(get_sata_drives) # Get list of SAS disks
# Checks whether it is a SAS disk
get_sas_drives() get_sas_drives()
{ {
gsas_smartdrives="" for drive in $Drive_list; do
for drive in $drives; do
gsas_smart_flag=$("$smartctl" -i "$drive" | grep -E "Transport protocol:[[:blank:]]+SAS" | awk '{print $3}') gsas_smart_flag=$("$smartctl" -i "$drive" | grep -E "Transport protocol:[[:blank:]]+SAS" | awk '{print $3}')
if [ "$gsas_smart_flag" = "SAS" ]; then if [ "$gsas_smart_flag" = "SAS" ]; then
gsas_smartdrives="$gsas_smartdrives $drive" SAS_list="$SAS_list $drive"
SAS_count=$((SAS_count + 1))
fi fi
done done
echo "$gsas_smartdrives"
} }
sasdrives=$(get_sas_drives) ### Fetch drive lists ###
get_smart_drives
# end of method 3. get_sata_drives
get_sas_drives
### Set email headers ### ### Set email headers ###
printf "%s\n" "To: ${email} printf "%s\n" "To: ${email}
@@ -100,158 +79,159 @@ Content-Transfer-Encoding: 7bit
Content-Disposition: inline Content-Disposition: inline
<html><head></head><body><pre style=\"font-size:14px; white-space:pre\">" > ${logfile} <html><head></head><body><pre style=\"font-size:14px; white-space:pre\">" > ${logfile}
if [ $Drive_count -eq 0 ]; then
echo "##### No SMART-enabled disks found on this system #####" >> "$logfile"
fi
###### summary sata ###### ###### Summary for SATA drives ######
( if [ $SATA_count -gt 0 ]; then
echo "########## SMART status report summary for all SATA drives on server ${freenashost} ##########"
echo ""
echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+--------------+-----+"
echo "|Device|Serial |Temp|Power|Start|Spin |ReAlloc|Current|Offline |Seek |Total |High |Command |Last |"
echo "| |Number | |On |Stop |Retry|Sectors|Pending|Uncorrec|Errors|Seeks |Fly |Timeout |Test |"
echo "| | | |Hours|Count|Count| |Sectors|Sectors | | |Writes|Count |Age |"
echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+--------------+-----+"
) >> "$logfile"
###### for each SATA drive ######
for drive in $satadrives; do
( (
devid=$(basename "$drive") echo "########## SMART status report summary for all SATA drives on server ${freenashost} ##########"
lastTestHours=$("$smartctl" -l selftest "$drive" | grep "# 1" | awk '{print $9}') echo ""
"$smartctl" -A -i -v 7,hex48 "$drive" | \ echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+-------+----+"
awk -v device="$devid" -v tempWarn="$tempWarn" -v tempCrit="$tempCrit" -v sectorsCrit="$sectorsCrit" \ echo "|Device|Serial |Temp|Power|Start|Spin |ReAlloc|Current|Offline |Seek |Total |High |Command|Last|"
-v testAgeWarn="$testAgeWarn" -v warnSymbol="$warnSymbol" -v critSymbol="$critSymbol" \ echo "| |Number | |On |Stop |Retry|Sectors|Pending|Uncorrec|Errors|Seeks |Fly |Timeout|Test|"
-v lastTestHours="$lastTestHours" ' echo "| | | |Hours|Count|Count| |Sectors|Sectors | | |Writes|Count |Age |"
/Serial Number:/{serial=$3} echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+-------+----+"
/190 Airflow_Temperature/{temp=$10} ) >> "$logfile"
/194 Temperature/{temp=$10}
/Power_On_Hours/{split($10,a,"+");sub(/h/,"",a[1]);onHours=a[1];} ###### Detail information for each SATA drive ######
/Start_Stop_Count/{startStop=$10} for drive in $SATA_list; do
/Spin_Retry_Count/{spinRetry=$10} (
/Reallocated_Sector/{reAlloc=$10} devid=$(basename "$drive")
/Current_Pending_Sector/{pending=$10} lastTestHours=$("$smartctl" -l selftest "$drive" | grep "# 1" | awk '{print $9}')
/Offline_Uncorrectable/{offlineUnc=$10} "$smartctl" -A -i -v 7,hex48 "$drive" | \
/Seek_Error_Rate/{seekErrors=("0x" substr($10,3,4));totalSeeks=("0x" substr($10,7))} awk -v device="$devid" -v tempWarn="$tempWarn" -v tempCrit="$tempCrit" -v sectorsCrit="$sectorsCrit" \
/High_Fly_Writes/{hiFlyWr=$10} -v testAgeWarn="$testAgeWarn" -v warnSymbol="$warnSymbol" -v critSymbol="$critSymbol" \
/Command_Timeout/{cmdTimeout=$10} -v lastTestHours="$lastTestHours" '
END { /Serial Number:/{serial=$3}
testAge=sprintf("%.0f", (onHours - lastTestHours) / 24); /190 Airflow_Temperature/{temp=$10}
if (temp > tempCrit || reAlloc > sectorsCrit || pending > sectorsCrit || offlineUnc > sectorsCrit) /194 Temperature/{temp=$10}
/Power_On_Hours/{split($10,a,"+");sub(/h/,"",a[1]);onHours=a[1];}
/Start_Stop_Count/{startStop=$10}
/Spin_Retry_Count/{spinRetry=$10}
/Reallocated_Sector/{reAlloc=$10}
/Current_Pending_Sector/{pending=$10}
/Offline_Uncorrectable/{offlineUnc=$10}
/Seek_Error_Rate/{seekErrors=("0x" substr($10,3,4));totalSeeks=("0x" substr($10,7))}
/High_Fly_Writes/{hiFlyWr=$10}
/Command_Timeout/{cmdTimeout=$10}
END {
testAge=sprintf("%.0f", (onHours - lastTestHours) / 24);
if (temp > tempCrit || reAlloc > sectorsCrit || pending > sectorsCrit || offlineUnc > sectorsCrit)
device=device " " critSymbol;
else if (temp > tempWarn || reAlloc > 0 || pending > 0 || offlineUnc > 0 || testAge > testAgeWarn)
device=device " " warnSymbol;
seekErrors=sprintf("%d", seekErrors);
totalSeeks=sprintf("%d", totalSeeks);
if (totalSeeks == "0") {
seekErrors="N/A";
totalSeeks="N/A";
}
if (temp > tempWarn || temp > tempCrit) temp=temp"*"
if (reAlloc > 0 || reAlloc > sectorsCrit) reAlloc=reAlloc"*"
if (pending > 0 || pending > sectorsCrit) pending=pending"*"
if (offlineUnc > 0 || offlineUnc > sectorsCrit) offlineUnc=offlineUnc"*"
if (testAge > testAgeWarn) testAge=testAge"*"
if (hiFlyWr == "") hiFlyWr="N/A";
if (cmdTimeout == "") cmdTimeout="N/A";
printf "|%-6s|%-24s|%-4s|%5s|%5s|%5s|%7s|%7s|%8s|%6s|%10s|%6s|%7s|%4s|\n",
device, serial, temp, onHours, startStop, spinRetry, reAlloc, pending, offlineUnc,
seekErrors, totalSeeks, hiFlyWr, cmdTimeout, testAge;
}'
) >> "$logfile"
done
(
echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+-------+----+"
) >> "$logfile"
fi
###### Summary for SAS drives ######
if [ $SAS_count -gt 0 ]; then
(
if [ $SATA_count -gt 0 ]; then
echo ""
fi
echo "########## SMART status report summary for all SAS drives on server ${freenashost} ##########"
echo ""
echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
echo "|Device|Serial |Temp|Start|Load |Defect|Uncorr|Uncorr|Uncorr|Non |"
echo "| |Number | |Stop |Unload|List |Read |Write |Verify|Medium|"
echo "| | | |Count|Count |Elems |Errors|Errors|Errors|Errors|"
echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
) >> "$logfile"
###### Detail information for each SAS drive ######
for drive in $SAS_list; do
(
devid=$(basename "$drive")
"$smartctl" -a "$drive" | \
awk -v device="$devid" -v tempWarn="$tempWarn" -v tempCrit="$tempCrit" \
-v warnSymbol="$warnSymbol" -v critSymbol="$critSymbol" '\
/Serial number:/{serial=$3}
/Current Drive Temperature:/{temp=$4} \
/start-stop cycles:/{startStop=$4} \
/load-unload cycles:/{loadUnload=$4} \
/grown defect list:/{defectList=$6} \
/read:/{readErrors=$8} \
/write:/{writeErrors=$8} \
/verify:/{verifyErrors=$8} \
/Non-medium error count:/{nonMediumErrors=$4} \
END {
if (temp > tempCrit)
device=device " " critSymbol; device=device " " critSymbol;
else if (temp > tempWarn || reAlloc > 0 || pending > 0 || offlineUnc > 0 || testAge > testAgeWarn) else if (temp > tempWarn)
device=device " " warnSymbol; device=device " " warnSymbol;
seekErrors=sprintf("%d", seekErrors); printf "|%-6s|%-24s| %3s|%5s|%6s|%6s|%6s|%6s|%6s|%6s|\n",
totalSeeks=sprintf("%d", totalSeeks); device, serial, temp, startStop, loadUnload, defectList, \
if (totalSeeks == "0") { readErrors, writeErrors, verifyErrors, nonMediumErrors;
seekErrors="N/A"; }'
totalSeeks="N/A"; ) >> "$logfile"
} done
if (temp > tempWarn || temp > tempCrit)
temp=temp"*"
else
temp=temp" "
if (reAlloc > 0 || reAlloc > sectorsCrit)
reAlloc=reAlloc"*"
if (pending > 0 || pending > sectorsCrit)
pending=pending"*"
if (offlineUnc > 0 || offlineUnc > sectorsCrit)
offlineUnc=offlineUnc"*"
if (testAge > testAgeWarn)
testAge=testAge"*"
if (hiFlyWr == "") hiFlyWr="N/A";
if (cmdTimeout == "") cmdTimeout="N/A";
printf "|%-6s|%-24s| %3s|%5s|%5s|%5s|%7s|%7s|%8s|%6s|%10s|%6s|%14s|%5s|\n",
device, serial, temp, onHours, startStop, spinRetry, reAlloc, pending, offlineUnc,
seekErrors, totalSeeks, hiFlyWr, cmdTimeout, testAge;
}'
) >> "$logfile"
done
(
echo "+------+------------------------+----+-----+-----+-----+-------+-------+--------+------+----------+------+--------------+-----+"
) >> "$logfile"
(
echo ""
echo ""
echo ""
) >> "$logfile"
###### summary sas ######
(
echo ""
echo "########## SMART status report summary for all SAS drives on server ${freenashost} ##########"
echo ""
echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
echo "|Device|Serial |Temp|Start|Load |Defect|Uncorr|Uncorr|Uncorr|Non |"
echo "| |Number | |Stop |Unload|List |Read |Write |Verify|Medium|"
echo "| | | |Count|Count |Elems |Errors|Errors|Errors|Errors|"
echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
) >> "$logfile"
###### for each SAS drive ######
for drive in $sasdrives; do
( (
devid=$(basename "$drive") echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
"$smartctl" -a "$drive" | \
awk -v device="$devid" -v tempWarn="$tempWarn" -v tempCrit="$tempCrit" \
-v warnSymbol="$warnSymbol" -v critSymbol="$critSymbol" '\
/Serial number:/{serial=$3}
/Current Drive Temperature:/{temp=$4} \
/start-stop cycles:/{startStop=$4} \
/load-unload cycles:/{loadUnload=$4} \
/grown defect list:/{defectList=$6} \
/read:/{readErrors=$8} \
/write:/{writeErrors=$8} \
/verify:/{verifyErrors=$8} \
/Non-medium error count:/{nonMediumErrors=$4} \
END {
if (temp > tempCrit)
device=device " " critSymbol;
else if (temp > tempWarn)
device=device " " warnSymbol;
printf "|%-6s|%-24s| %3s|%5s|%6s|%6s|%6s|%6s|%6s|%6s|\n",
device, serial, temp, startStop, loadUnload, defectList, \
readErrors, writeErrors, verifyErrors, nonMediumErrors;
}'
) >> "$logfile" ) >> "$logfile"
done fi
(
echo "+------+------------------------+----+-----+------+------+------+------+------+------+"
) >> "$logfile"
if [ $SATA_count -gt 0 ] || [ $SAS_count -gt 0 ]; then
###### for SATA drives ######
for drive in $satadrives; do ###### Emit SATA drive information ######
brand=$("$smartctl" -i "$drive" | grep "Model Family" | sed "s/^.* //") for drive in $SATA_list; do
if [ -z "$brand" ]; then brand=$("$smartctl" -i "$drive" | grep "Model Family" | awk '{print $3, $4, $5, $6, $7}')
brand=$("$smartctl" -i "$drive" | grep "Device Model" | sed "s/^.* //") if [ -z "$brand" ]; then
fi brand=$("$smartctl" -i "$drive" | grep "Device Model" | awk '{print $3, $4, $5, $6, $7}')
serial=$("$smartctl" -i "$drive" | grep "Serial Number" | sed "s/^.* //") fi
( serial=$("$smartctl" -i "$drive" | grep "Serial Number" | awk '{print $3}')
echo "" (
echo "########## SMART status report for $drive drive (${brand} : ${serial}) ##########" echo ""
"$smartctl" -n never -H -A -l error "$drive" echo "########## SMART status for SATA drive $drive $serial (${brand}) ##########"
"$smartctl" -n never -l selftest "$drive" | grep "# 1 \\|Num" | cut -c6- "$smartctl" -n never -H -A -l error "$drive"
) >> "$logfile" "$smartctl" -n never -l selftest "$drive" | grep "# 1 \\|Num" | cut -c6-
done ) >> "$logfile"
done
###### for SAS drives ######
for drive in $sasdrives; do ###### Emit SAS drive information ######
devid=$(basename "$drive") for drive in $SAS_list; do
brand=$("$smartctl" -i "$drive" | grep "Product" | sed "s/^.* //") devid=$(basename "$drive")
serial=$("$smartctl" -i "$drive" | grep "Serial number" | sed "s/^.* //") brand=$("$smartctl" -i "$drive" | grep "Product" | sed "s/^.* //")
( serial=$("$smartctl" -i "$drive" | grep "Serial number" | sed "s/^.* //")
echo "" (
echo "########## SMART status report for $drive drive (${brand} : ${serial}) ##########" echo ""
"$smartctl" -n never -H -A -l error "$drive" echo "########## SMART status for SAS drive $drive $serial (${brand}) ##########"
"$smartctl" -n never -l selftest "$drive" | grep "# 1 \\|Num" | cut -c6- "$smartctl" -n never -H -A -l error "$drive"
) >> "$logfile" "$smartctl" -n never -l selftest "$drive" | grep "# 1 \\|Num" | cut -c6-
done ) >> "$logfile"
done
fi
sed -i '' -e '/smartctl 7.*/d' "$logfile" sed -i '' -e '/smartctl 7.*/d' "$logfile"
sed -i '' -e '/smartctl 6.*/d' "$logfile" sed -i '' -e '/smartctl 6.*/d' "$logfile"