[techsupport] Added a lock to avoid running techsupport in parallel (sonic-net#2065)

- What I did
Added logic to generate_dump script to avoid parallel execution of techsupport.
If a second instance of techsupport starts when one is already running, the second one exits with an appropriate error code

- Why I did it
1. Running multiple dumps in paralell has no real use case
2. High CPU load
3. saisdkdump is not designed to run in paralell. When run, these sort of logs are seen indicating failure.

Signed-off-by: Vivek Reddy Karri <[email protected]>

Author: vivekrnv

scripts/generate_dump

@@ -6,9 +6,14 @@
 
 set -u
 
-ERROR_TAR_FAILED=5
-ERROR_PROCFS_SAVE_FAILED=6
-ERROR_INVALID_ARGUMENT=10
+EXT_SUCCESS=0
+EXT_GENERAL=1
+EXT_LOCKFAIL=2
+EXT_RECVSIG=3
+EXT_RETRY=4
+EXT_TAR_FAILED=5
+EXT_PROCFS_SAVE_FAILED=6
+EXT_INVALID_ARGUMENT=10
 
 TAR=tar
 MKDIR=mkdir
@@ -39,21 +44,38 @@ USER=${USER:-root}
 TIMEOUT_MIN="5"
 SKIP_BCMCMD=0
 SAVE_STDERR=true
-RETURN_CODE=0
+RETURN_CODE=$EXT_SUCCESS
 DEBUG_DUMP=false
 
+# lock dirs/files
+LOCKDIR="/tmp/techsupport-lock"
+PIDFILE="${LOCKDIR}/PID"
+
+# Remove lock directory and exit, let user decide if they want to retry
+rm_lock_and_exit()
+{
+    $RM $V -rf ${LOCKDIR}
+    exit $EXT_RETRY
+}
+
+handle_exit()
+{
+    ECODE=$?
+    echo "Removing lock. Exit: $ECODE" >&2
+    $RM $V -rf ${LOCKDIR}
+}
+
 handle_signal()
 {
     echo "Generate Dump received interrupt" >&2
     $RM $V -rf $TARDIR
-    exit 1
+    exit $EXT_RECVSIG
 }
-trap 'handle_signal' SIGINT
 
 handle_error() {
     if [ "$1" != "0" ]; then
         echo "ERR: RC:-$1 observed on line $2" >&2
-        RETURN_CODE=1
+        RETURN_CODE=$EXT_GENERAL
     fi
 }
 
@@ -103,7 +125,7 @@ save_bcmcmd() {
         filepath="${filepath}.gz"
     fi
     ($TAR $V -rhf $TARFILE -C $DUMPDIR "$tarpath" \
-        || abort "${ERROR_TAR_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
+        || abort "${EXT_TAR_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
         && $RM $V -rf "$filepath"
     end_t=$(date +%s%3N)
     echo "[ save_bcmcmd:$cmd ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
@@ -213,7 +235,7 @@ save_cmd() {
     fi
 
     ($TAR $V -rhf $TARFILE -C $DUMPDIR "$tarpath" \
-        || abort "${ERROR_TAR_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
+        || abort "${EXT_TAR_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
         && $RM $V -rf "$filepath"
     end_t=$(date +%s%3N)
     echo "[ save_cmd:$cmd ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
@@ -766,7 +788,7 @@ save_file() {
 
     if $do_tar_append; then
         ($TAR $V -rhf $TARFILE -C $DUMPDIR "$tar_path" \
-            || abort "${ERROR_PROCFS_SAVE_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
+            || abort "${EXT_PROCFS_SAVE_FAILED}" "tar append operation failed. Aborting to prevent data loss.") \
             && $RM $V -f "$gz_path"
     fi
     end_t=$(date +%s%3N)
@@ -975,7 +997,7 @@ save_log_files() {
 
     # Append the log folder to the main tarball
     ($TAR $V -rhf $TARFILE -C $DUMPDIR ${BASE}/log \
-        || abort "${ERROR_TAR_FAILED}" "tar append operation failed. Aborting for safety") \
+        || abort "${EXT_TAR_FAILED}" "tar append operation failed. Aborting for safety") \
         && $RM $V -rf $TARDIR/log
     end_t=$(date +%s%3N)
     echo "[ TAR /var/log Files ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
@@ -1004,7 +1026,7 @@ save_warmboot_files() {
 
         ($TAR $V --warning=no-file-removed  -rhf $TARFILE -C $DUMPDIR --mode=+rw \
             $BASE/warmboot \
-            || abort "${ERROR_TAR_FAILED}" "Tar append operation failed. Aborting for safety.") \
+            || abort "${EXT_TAR_FAILED}" "Tar append operation failed. Aborting for safety.") \
             && $RM $V -rf $TARDIR
     fi
     end_t=$(date +%s%3N)
@@ -1144,11 +1166,6 @@ main() {
     trap 'handle_error $? $LINENO' ERR
     local start_t=0
     local end_t=0
-    if [ `whoami` != root ] && ! $NOOP;
-    then
-        echo "$0: must be run as root (or in sudo)" >&2
-        exit 10
-    fi
     NUM_ASICS=$(get_asic_count)
     ${CMD_PREFIX}renice +5 -p $$ >> /dev/null
     ${CMD_PREFIX}ionice -c 2 -n 5 -p $$ >> /dev/null
@@ -1174,7 +1191,7 @@ main() {
         /proc/softirqs /proc/stat /proc/swaps /proc/sysvipc /proc/timer_list \
         /proc/uptime /proc/version /proc/vmallocinfo /proc/vmstat \
         /proc/zoneinfo \
-        || abort "${ERROR_PROCFS_SAVE_FAILED}" "Proc saving operation failed. Aborting for safety."
+        || abort "${EXT_PROCFS_SAVE_FAILED}" "Proc saving operation failed. Aborting for safety."
     end_t=$(date +%s%3N)
     echo "[ Capture Proc State ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
 
@@ -1307,7 +1324,7 @@ main() {
         --exclude="*/etc/ssl/certs/*" \
         --exclude="*/etc/ssl/private/*" \
         $BASE/etc \
-        || abort "${ERROR_TAR_FAILED}" "Tar append operation failed. Aborting for safety.") \
+        || abort "${EXT_TAR_FAILED}" "Tar append operation failed. Aborting for safety.") \
         && $RM $V -rf $TARDIR
     end_t=$(date +%s%3N)
     echo "[ TAR /etc Files ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
@@ -1473,7 +1490,7 @@ while getopts ":xnvhzas:t:r:d" opt; do
             ;;
         h)
             usage
-            exit 0
+            exit $EXT_SUCCESS
             ;;
         v)
             # echo commands about to be run to stderr
@@ -1501,7 +1518,7 @@ while getopts ":xnvhzas:t:r:d" opt; do
         s)
             SINCE_DATE="${OPTARG}"
             # validate date expression
-            date --date="${SINCE_DATE}" &> /dev/null || abort "${ERROR_INVALID_ARGUMENT}" "Invalid date expression passed: '${SINCE_DATE}'"
+            date --date="${SINCE_DATE}" &> /dev/null || abort "${EXT_INVALID_ARGUMENT}" "Invalid date expression passed: '${SINCE_DATE}'"
             ;;
         t)
             TIMEOUT_MIN="${OPTARG}"
@@ -1514,9 +1531,47 @@ while getopts ":xnvhzas:t:r:d" opt; do
             ;;
         /?)
             echo "Invalid option: -$OPTARG" >&2
-            exit 1
+            exit $EXT_GENERAL
             ;;
     esac
 done
 
-main
+# Check permissions before proceeding further
+if [ `whoami` != root ] && ! $NOOP;
+then
+    echo "$0: must be run as root (or in sudo)" >&2
+    exit $EXT_INVALID_ARGUMENT
+fi
+
+##
+## Attempt Locking
+##
+
+if mkdir "${LOCKDIR}" &>/dev/null; then
+    trap 'handle_exit' EXIT
+    echo "$$" > "${PIDFILE}"
+    # This handler will exit the script upon receiving these interrupts
+    # Trap configured on EXIT will be triggered by the exit from handle_signal function
+    trap 'handle_signal' SIGINT SIGHUP SIGQUIT SIGTERM
+    echo "Lock succesfully accquired and installed signal handlers"
+    # Proceed with the actual code
+    main
+else
+    # lock failed, check if the other PID is alive
+    PID_PROG="$(cat "${PIDFILE}")"
+
+    if [ $? != 0 ]; then
+        # Another instance is probably about to remove the lock or PIDfile doesn't exist
+        rm_lock_and_exit
+    fi
+
+    if ! kill -0 $PID_PROG &>/dev/null; then
+        # Lock is stale
+        echo "Removing stale lock of nonexistant PID ${PID_PROG}"
+        rm_lock_and_exit
+    else
+        # Lock is valid and the other instance is active. Exit Now
+        echo "Accquiring lock failed, PID ${PID_PROG} is active" >&2
+        exit $EXT_LOCKFAIL
+    fi
+fi