Change leasing LEASE_PID var to LEASE_ID (it doesn't need to be a PID)

Author: Tom Barnes

src/integration-tests/bash/lease.sh

@@ -29,33 +29,35 @@ cat << EOF
 
   Usage:
  
-    Obtain       : lease.sh -o pid [-t timeout]
-    Renew        : lease.sh -r pid  
-    Delete       : lease.sh -d pid 
+    Obtain       : lease.sh -o id [-t timeout]
+    Renew        : lease.sh -r id  
+    Delete       : lease.sh -d id 
     Force Delete : lease.sh -f 
     Show         : lease.sh -s
     Help         : lease.sh -h
 
     The user is expected to call this script multiple times throughout
     the ownership of a lease:
  
-    1) First, call 'lease.sh -o pid' to try obtain a lease.  If 
+    1) First, call 'lease.sh -o id' to try obtain a lease.  If 
        this fails, fail your script, since failure may mean some other
        owner has ownership.  This call will block up to 30 minutes
        while trying to get the lease (tunable via -t), and will grab
        the lease if no other process owns it or no other process has touched
-       it for a period of 10 minutes.
+       it for a period of 10 minutes.  'id' should be some unique
+       string suitable for use as a filename - a pid is sufficient
+       or a `uuidgen` value.
  
-    2) While running, call 'lease.sh -r pid' every few minutes
+    2) While running, call 'lease.sh -r id' every few minutes
        to renew the lease and retain ownership.   Fail if this fails,
        since that means you may have lost the lease.  Be careful to call
        this within 10 minutes of last obtaining or renewing the lease, 
        otherwise the lease expires and another process is free to obtain
-       ownership.  Use the same value of 'pid' as was specified in step (1).
+       ownership.  Use the same value of 'id' as was specified in step (1).
   
-    3) Finally, call 'lease.sh -d pid' to delete and give up your
+    3) Finally, call 'lease.sh -d id' to delete and give up your
        lease.  This will fail if you are not the current owner.  Use the same
-       value of 'pid' as was specified in step (1).  If you forget to 
+       value of 'id' as was specified in step (1).  If you forget to 
        call it, no one else will be able to get the lease for up to 10
        minutes.
  
@@ -120,8 +122,8 @@ function main {
   # Save command line for error messages...
   COMMAND_LINE="$0 $*"
 
-  # Local PID that owns the K8S lease.  This is passed in on the command line.
-  LOCAL_PID=""
+  # Local ID that owns the K8S lease.  This is passed in on the command line.
+  LOCAL_ID=""
 
   local mode=""
 
@@ -137,23 +139,23 @@ function main {
         ;;
     -o) if [ "$mode" = "" ] && [ ! "$2" = "" ]; then
           mode="obtainLease"
-          LOCAL_PID="${2}"
+          LOCAL_ID="${2}"
           shift
         else
           mode="commandLineError"
         fi
         ;;
     -r) if [ "$mode" = "" ] && [ ! "$2" = "" ]; then
           mode="renewLease"
-          LOCAL_PID="${2}"
+          LOCAL_ID="${2}"
           shift
         else
           mode="commandLineError"
         fi
         ;;
     -d) if [ "$mode" = "" ] && [ ! "$2" = "" ]; then
           mode="deleteRemoteLeaseSafe"
-          LOCAL_PID="${2}"
+          LOCAL_ID="${2}"
           shift
         else
           mode="commandLineError"
@@ -191,7 +193,7 @@ function main {
     shift
   done
 
-  LOCAL_ROOT="${LOCAL_ROOT}/lease/pid-${LOCAL_PID}"
+  LOCAL_ROOT="${LOCAL_ROOT}/lease/id-${LOCAL_ID}"
 
   # Execute command line
 
@@ -240,7 +242,7 @@ function getLocalLease {
 function makeLocalLease {
 #
 # Make a candidate lease file in ${LOCAL_ROOT}/[expired/]${LOCAL_FILE} 
-# Embed pid, hostname, timestamp, etc to make it easy to identify the owner.
+# Embed id, hostname, timestamp, etc to make it easy to identify the owner.
 # returns non-zero if fails
 # Usage: Pass "expired" as $1 to create a lease that immediately expires
 #        (its timestamp value will be 0)  This file will be located
@@ -262,7 +264,7 @@ function makeLocalLease {
 timestamp=0
 date=`date '+%m-%d-%YT%H:%M:%S'`
 host=$HOST
-pid=$LOCAL_PID
+id=$LOCAL_ID
 user=$USER
 EOF
   else
@@ -271,7 +273,7 @@ EOF
 timestamp=`date +%s`
 date=`date '+%m-%d-%YT%H:%M:%S'`
 host=$HOST
-pid=$LOCAL_PID
+id=$LOCAL_ID
 user=$USER
 EOF
   fi

src/integration-tests/bash/run.sh

@@ -101,16 +101,17 @@
 #   BRANCH_NAME    Git branch name.
 #                  Default is determined by calling 'git branch'.
 #
-#   LEASE_PID      Set to a PID to (A) periodically renew a lease on
+#   LEASE_ID       Set to a unique value to (A) periodically renew a lease on
 #                  the k8s cluster that indicates that no other
 #                  run.sh should attempt to use the cluster, and (B)
 #                  delete this lease when run.sh completes.
 #                  If "true" the caller must previously
 #                  obtain the lease external to the run.sh
-#                  using "lease.sh -o $LEASE_PID", where PID corresponds
+#                  using "lease.sh -o $LEASE_ID", where ID corresponds
 #                  to a process that's expected to continue
 #                  throughout the run (typically the parent 
-#                  process).
+#                  process) or some uuid potentially generated by uuidgen
+#                  command or some-such..
 #
 # The following additional overrides are currently only used when
 # WERCKER=true:
@@ -205,31 +206,31 @@ $1
 }
 
 
-# Attempt to renew the k8s lease if $LEASE_PID is set.  This should be called
+# Attempt to renew the k8s lease if $LEASE_ID is set.  This should be called
 # every few minutes throughout the run, and is currently called at the start
-# of each test.   See LEASE_PID instructions above for instructions about
+# of each test.   See LEASE_ID instructions above for instructions about
 # how to obtain the lease prior to calling run.sh.
 function renewLease {
-  if [ ! "$LEASE_PID" = "" ]; then
+  if [ ! "$LEASE_ID" = "" ]; then
     # RESULT_DIR may not have been created yet, so use /tmp
     local outfile=/tmp/acc_test_renew_lease.out
-    $SCRIPTPATH/lease.sh -r $LEASE_PID > $outfile 2>&1
+    $SCRIPTPATH/lease.sh -r "$LEASE_ID" > $outfile 2>&1
     if [ $? -ne 0 ]; then
       trace "Lease renew error:"
       echo "" >> $outfile
-      echo "ERROR: Could not renew lease on k8s cluster for LEASE_PID=$LEASE_PID." >> $outfile
-      echo "Used '$SCRIPTPATH/lease.sh -r $LEASE_PID' to try renew the lease." >> $outfile
+      echo "ERROR: Could not renew lease on k8s cluster for LEASE_ID=\"$LEASE_ID\"." >> $outfile
+      echo "Used '$SCRIPTPATH/lease.sh -r \"$LEASE_ID\"' to try renew the lease." >> $outfile
       echo "Some of the potential reasons for this failure are that another run" >> $outfile
       echo "may have obtained the lease, the lease may have been externally " >> $outfile
       echo "deleted, or the caller of run.sh may have forgotten to obtain the" >> $outfile
-      echo "lease before calling run.sh (using 'lease.sh -o $LEASE_PID'). " >> $outfile
+      echo "lease before calling run.sh (using 'lease.sh -o \"$LEASE_ID\"'). " >> $outfile
       echo "To force delete a lease no matter who owns the lease," >> $outfile
       echo "call 'lease.sh -f' or 'kubernetes delete cm acceptance-test-lease'" >> $outfile
       echo "(this should only be done when sure there's no current run.sh " >> $outfile
       echo "that owns the lease).  To view the current lease holder," >> $outfile
       echo "use 'lease.sh -s'.  To disable this lease check, do not set" >> $outfile
-      echo "the LEASE_PID environment variable.  For more information, see" >> $outfile
-      echo "LEASE_PID in the instructions embedded at the top of run.sh." >> $outfile
+      echo "the LEASE_ID environment variable.  For more information, see" >> $outfile
+      echo "LEASE_ID in the instructions embedded at the top of run.sh." >> $outfile
       echo "" >> $outfile
       cat $outfile
       fail "Could not renew lease on k8s cluster"
@@ -383,7 +384,7 @@ function state_dump {
      local PV_ROOT="`cat /tmp/test_suite.pv_root`"
      local PROJECT_ROOT="`cat /tmp/test_suite.project_root`"
      local SCRIPTPATH="$PROJECT_ROOT/src/integration-tests/bash"
-     local LEASE_PID="`cat /tmp/test_suite.lease_pid`"
+     local LEASE_ID="`cat /tmp/test_suite.lease_id`"
 
      if [ ! -d "$RESULT_DIR" ]; then
         trace State dump exiting early.  RESULT_DIR \"$RESULT_DIR\" does not exist or is not a directory.
@@ -439,9 +440,9 @@ function state_dump {
      trace Job failed.  See ${outfile}.
   fi
 
-  if [ ! "$LEASE_PID" = "" ]; then
+  if [ ! "$LEASE_ID" = "" ]; then
     # release the lease if we own it
-    ${SCRIPTPATH}/lease.sh -d $LEASE_PID > ${RESULT_DIR}/release_lease.out 2>&1
+    ${SCRIPTPATH}/lease.sh -d "$LEASE_ID" > ${RESULT_DIR}/release_lease.out 2>&1
     if [ "$?" = "0" ]; then
       trace Lease released.
     else
@@ -2448,7 +2449,7 @@ function test_suite_init {
                    WEBLOGIC_IMAGE_PULL_SECRET_NAME \
                    WERCKER \
                    JENKINS \
-                   LEASE_PID;
+                   LEASE_ID;
     do
       trace "Customizable env var before: $varname=${!varname}"
     done
@@ -2464,7 +2465,7 @@ function test_suite_init {
       [ ! "$?" = "0" ] && fail "Error: Could not determine branch.  Run script from within a git repo".
     fi
 
-    export LEASE_PID=${LEASE_PID}
+    export LEASE_ID="${LEASE_ID}"
 
     # The following customizable exports are currently only customized by WERCKER
     export IMAGE_TAG_OPERATOR=${IMAGE_TAG_OPERATOR:-`echo "test_${BRANCH_NAME}" | sed "s#/#_#g"`}
@@ -2490,7 +2491,7 @@ function test_suite_init {
                    WEBLOGIC_IMAGE_PULL_SECRET_NAME \
                    WERCKER \
                    JENKINS \
-                   LEASE_PID;
+                   LEASE_ID;
     do
       trace "Customizable env var after: $varname=${!varname}"
     done
@@ -2513,7 +2514,7 @@ function test_suite_init {
     echo "${RESULT_ROOT}" > /tmp/test_suite.result_root
     echo "${PV_ROOT}" > /tmp/test_suite.pv_root
     echo "${PROJECT_ROOT}" > /tmp/test_suite.project_root
-    echo "${LEASE_PID}" > /tmp/test_suite.lease_pid
+    echo "${LEASE_ID}" > /tmp/test_suite.lease_id
 
     # Declare we're in a test.  We did not declare at the start
     # of the function to ensure that any env vars that might

wercker.yml

@@ -113,22 +113,26 @@ integration-test:
       kubectl version
 
       # obtain an exclusive k8s cluster lease using the 'lease.sh' helper script
-      #   - first set LEASE_PID to a PID
+      #   - first set LEASE_ID to a unique value
       #   - then try obtain the lease, block up to 100 minutes (wercker pipeline should timeout before then)
-      #   - finally, run.sh will periodically try renew the lease as it runs (using $LEASE_PID)
+      #   - finally, run.sh will periodically try renew the lease as it runs (using $LEASE_ID)
       #   - if run.sh fails when it tries to renew the lease (as something else took it, etc), it will exit early
       #   - when run.sh exits, it will try release the lease if it's still the owner...
-      export LEASE_PID=$$
+      export LEASE_ID="`uuidgen`-pid$$"
+      echo @@
+      echo @@ "Obtaining lease!"
       echo @@
       echo @@ "About to block up to the 100 minutes trying to get exclusive access to the kubernetes cluster."
       echo @@ "If this blocks unexpectedly and you are sure that the kubernetes cluster isn't in use by "
       echo @@ "another Wercker pipeline, you can force the lease to free up via 'kubectl delete cm acceptance-test-lease'."
-      echo @@ "See LEASE_PID in run.sh for details about this heuristic."
-      echo @@ "PID=$LEASE_PID host=$HOST date=`date` user=$USER."
+      echo @@ "See LEASE_ID in run.sh for details about this heuristic."
+      echo @@ "LEASE_ID=$LEASE_ID host=$HOST date=`date` user=$USER."
+      echo @@
       echo @@ "Current lease owner (if any):"
       $WERCKER_SOURCE_DIR/src/integration-tests/bash/lease.sh -s 
+      echo @@
       echo @@ "About to try obtain lease:"
-      $WERCKER_SOURCE_DIR/src/integration-tests/bash/lease.sh -o $LEASE_PID -t $((100 * 60))
+      $WERCKER_SOURCE_DIR/src/integration-tests/bash/lease.sh -o "$LEASE_ID" -t $((100 * 60))
       echo @@
 
       # create pull secrets
@@ -182,7 +186,7 @@ integration-test:
       # release lease in case run.sh failed to release it
       # (the following command only releases the release after confirming this pipeline still owns it)
       # TBD Calling this somehow seems to fail the wercker run
-      # $WERCKER_SOURCE_DIR/src/integration-tests/bash/lease.sh -d $LEASE_PID > /tmp/junk 2>&1
+      # $WERCKER_SOURCE_DIR/src/integration-tests/bash/lease.sh -d "$LEASE_ID" > /tmp/junk 2>&1
 
       # clean up
       yum clean all