b=11171

[fs/lustre-release.git] / lustre-iokit / sgpdd-survey / sgpdd-survey

diff --git a/lustre-iokit/sgpdd-survey/sgpdd-survey b/lustre-iokit/sgpdd-survey/sgpdd-survey
index b5c1f01..94946c0 100755 (executable)
--- a/lustre-iokit/sgpdd-survey/sgpdd-survey
+++ b/lustre-iokit/sgpdd-survey/sgpdd-survey
@@ -11,24 +11,26 @@
 
 # result file prefix.  date/time+hostname makes unique
 # NB ensure the path exists if it includes subdirs
-rslt=${rslt:-"/tmp/sgpdd_survey_`date +%F@%R`_`uname -n`"}
+rslt_loc=${rslt_loc:-"/tmp"}
+rslt=${rslt:-"$rslt_loc/sgpdd_survey_`date +%F@%R`_`uname -n`"}
 
 # what to do (read or write)
-actions="write read"
+actions=${actions:-"write read"}
 
 # total size per device (MBytes)
 # NB bigger than device cache is good
-size=8192
+size=${size:-8192}
 
 # record size (KBytes)
-rszlo=1024
-rszhi=1024
+rszlo=${rszlo:-1024}
+rszhi=${rszhi:-1024}
 
 # Concurrent regions per device
 crglo=${crglo:-1}
 crghi=${crghi:-256}
 
 # threads to share between concurrent regions per device
+# multiple threads per region simulates a deeper request queue
 # NB survey skips over #thr < #regions and #thr/#regions > SG_MAX_QUEUE
 thrlo=${thrlo:-1}
 thrhi=${thrhi:-4096}
@@ -36,8 +38,6 @@ thrhi=${thrhi:-4096}
 #####################################################################
 # leave the rest of this alone unless you know what you're doing...
 
-# sgp_dd's idea of disk sector size (Bytes)
-bs=512
 # and max # threads one instance will spawn
 SG_MAX_QUEUE=16
 
@@ -48,12 +48,12 @@ if [ "$scsidevs" ]; then
         # we will test for a LUN, the test for a partition
         # if the partition number is > 9 this will fail
     for d in $scsidevs; do
-       devs[$i]=`sg_map | awk "{if ($ 2 == \"$d\") print $ 1}"`
+       devs[$i]=`sg_map | awk "{if (\\\$2 == \"$d\") print \\\$1}"`
         if [ -z "${devs[i]}" ]; then
             echo "Can't find SG device for $d, testing for partition"
             pt=`echo $d | sed 's/[0-9]$//'`
             # Try again
-            devs[$i]=`sg_map | awk "{if ($ 2 == \"$pt\") print $ 1}"`
+            devs[$i]=`sg_map | awk "{if (\\\$2 == \"$pt\") print \\\$1}"`
             if [ -z "${devs[i]}" ]; then
                 echo "Can't find SG device $pt"
                 exit 1
@@ -79,6 +79,13 @@ fi
 
 ndevs=${#devs[@]}
 
+# determine block size. This should also work for raw devices
+# If it fails, set to 512
+bs=$((`sg_readcap -b ${devs[0]} | awk '{print $2}'`))
+if [ $bs == 0  ];then
+       echo "sg_readcap failed, setting block size to 512"
+       bs=512
+fi
 rsltf=${rslt}.summary
 workf=${rslt}.detail
 echo -n > $rsltf
@@ -94,6 +101,8 @@ print_summary () {
     echo $minusn "$*"
 }
 
+print_summary "$(date) sgpdd-survey on $rawdevs$scsidevs from $(hostname)"
+
 for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
     for ((crg=$crglo;crg<=$crghi;crg*=2)); do 
        for ((thr=$thrlo;thr<=$thrhi;thr*=2)); do
@@ -125,6 +134,11 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                t0=`date +%s.%N`
                for ((i=0;i<ndevs;i++)); do
                    dev=${devs[i]}
+                   devsize=$((bs*`sg_readcap -b ${dev} | awk '{print $1}'`/1024))
+                   if [ $devsize -lt $actual_size ]; then
+                       echo -e "device $(sg_map | grep $dev | awk '{ print $2; }') not big enough: $devsize < $actual_size.\nConsider reducing \$size"
+                       exit 1
+                   fi
                    if [ $action = read ]; then
                        inf="if=$dev"
                        outf="of=/dev/null"
@@ -148,7 +162,9 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                for ((i=0;i<ndevs;i++)); do
                    for ((j=0;j<crg;j++)); do
                        rtmp=${tmpf}_${i}_${j}
-                       if grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
+                       if grep 'error' $rtmp > /dev/null 2>&1; then
+                               echo "Error found in $rtmp"
+                       elif grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
                            ok=$((ok + 1))
                        fi
                        cat ${rtmp} >> $tmpf