X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=automate-default.sh;h=a51144a2afbe719e57dbee5afca7c44814b768bd;hb=5684d125af2ff554c960c402447cf871396108c6;hp=73dc1103e196119807b9b4c27a3921fe219d761b;hpb=0d3d555aa76958ef55d946b21028be976a84e113;p=monitor.git

diff --git a/automate-default.sh b/automate-default.sh
index 73dc110..a51144a 100755
--- a/automate-default.sh
+++ b/automate-default.sh
@@ -1,14 +1,19 @@
 #!/bin/bash
 
+export PATH=$PATH:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
+
 # NOTE: Must be an absolute path to guarantee it is read.
-source /usr/share/monitor-server/monitorconfig.py
+INSTALLPATH=/usr/share/monitor/
+$INSTALLPATH/commands/shconfig.py >  $INSTALLPATH/monitorconfig.sh
+source $INSTALLPATH/monitorconfig.sh
 cd ${MONITOR_SCRIPT_ROOT}
 set -e
 DATE=`date +%Y-%m-%d-%T`
-MONITOR_PID="$HOME/monitor/SKIP"
+MONITOR_PID="${MONITOR_SCRIPT_ROOT}/SKIP"
 
+echo "#######################################"; echo "Running Monitor at $DATE"; echo "######################################"
 echo "Performing API test"
-API=$(./testapi.py)
+API=$(${MONITOR_SCRIPT_ROOT}/tools/testapi.py)
 if [ "$API" != "ok" ] ; then 
 	# NOTE: Do not try to run any commands if the API is obviously broken.
 	echo "API IS DOWN : "`date`
@@ -20,7 +25,12 @@ if [ -f $MONITOR_PID ] ; then
 		echo "KILLING Monitor"
 		PID=`cat $MONITOR_PID`
 		rm -f $MONITOR_PID
-		${MONITOR_SCRIPT_ROOT}/kill.cmd.sh $PID
+		if [ -z $PID ] ; then
+			${MONITOR_SCRIPT_ROOT}/tools/kill.cmd.sh $PID
+			echo "done."
+		else
+			echo "No PID to be killed."
+		fi
 	else 
 		# skipping monitor
 		echo "SKIPPING Monitor"
@@ -29,65 +39,44 @@ if [ -f $MONITOR_PID ] ; then
 fi
 echo $$ > $MONITOR_PID
 
-echo "Performing Findbad Nodes"
-#########################
-# 1. FINDBAD NODES 
-rm -f ${MONITOR_DATA_ROOT}/production.findbad2.pkl
-${MONITOR_SCRIPT_ROOT}/findbad.py --increment --cachenodes --debug=0 --dbname="findbad2" $DATE || :
-cp ${MONITOR_DATA_ROOT}/production.findbad2.pkl ${MONITOR_DATA_ROOT}/production.findbad.pkl
-ps ax | grep BatchMode | grep -v grep | awk '{print $1}' | xargs kill || :
-
-echo "Performing Findbad PCUs"
-#########################
-# 2. FINDBAD PCUS
-rm -f ${MONITOR_DATA_ROOT}/production.findbadpcus2.pkl
-${MONITOR_SCRIPT_ROOT}/findbadpcu.py --increment --refresh --debug=0 --dbname=findbadpcus2 $DATE || :
-cp ${MONITOR_DATA_ROOT}/production.findbadpcus2.pkl ${MONITOR_DATA_ROOT}/production.findbadpcus.pkl
-# clean up stray 'locfg' processes that hang around inappropriately...
-ps ax | grep locfg | grep -v grep | awk '{print $1}' | xargs kill || :
+# SETUP act_all database if it's not there.
+if [ ! -f ${MONITOR_SCRIPT_ROOT}/actallsetup.flag ]; then
+	if ! python -c 'import database; database.dbLoad("act_all")' 2>/dev/null ; then 
+		touch ${MONITOR_SCRIPT_ROOT}/actallsetup.flag
+	fi
+fi
 
-echo "Generating web data"
-# badcsv.txt
-${MONITOR_SCRIPT_ROOT}/printbadcsv.py  | grep -v loading | tr -d ' ' > badcsv.txt
-cp badcsv.txt /plc/data/var/www/html/monitor/
-${MONITOR_SCRIPT_ROOT}/showlatlon.py | head -9 | awk 'BEGIN {print "<table>"} { print "<tr><td>", $0, "</td></tr>"} END{print "</table>"}'  | sed -e 's\|\</td><td>\g' > /plc/data/var/www/html/monitor/regions.html
 
-echo "Performing uptime changes for sites, nodes, and pcus"
-########################
-# 3. record last-changed for sites, nodes and pcus.
-${MONITOR_SCRIPT_ROOT}/sitebad.py --increment || :
-${MONITOR_SCRIPT_ROOT}/nodebad.py --increment || :
-${MONITOR_SCRIPT_ROOT}/pcubad.py --increment || :
+set +e
+AGENT=`ps ax | grep ssh-agent | grep -v grep`
+set -e
+if [ -z "$AGENT" ] ; then
+        echo "starting ssh agent"
+        # if no agent is running, set it up.
+        ssh-agent > ${MONITOR_SCRIPT_ROOT}/agent.sh
+        source ${MONITOR_SCRIPT_ROOT}/agent.sh
+        ssh-add /etc/planetlab/myops_ssh_key.rsa
+        ssh-add /etc/planetlab/debug_ssh_key.rsa
+        ssh-add /etc/planetlab/root_ssh_key.rsa
+fi
+#TODO: should add a call to ssh-add -l to check if the keys are loaded or not.
+source ${MONITOR_SCRIPT_ROOT}/agent.sh
 
-echo "Converting pkl files to phpserial"
-#########################
-# 4. convert pkl to php serialize format.
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i findbadpcus2 -o findbadpcus
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i act_all -o act_all
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i plcdb_hn2lb -o plcdb_hn2lb
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i findbad -o findbadnodes
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i ad_dbTickets -o ad_dbTickets
-${MONITOR_SCRIPT_ROOT}/pkl2php.py -i idTickets -o idTickets
+${MONITOR_SCRIPT_ROOT}/commands/syncwithplc.py $DATE || :
+service plc restart monitor
 
-echo "Archiving pkl files"
+echo "Performing FindAll Nodes"
 #########################
-# Archive pkl files.
-for f in findbad act_all findbadpcus l_plcnodes site_persistflags node_persistflags pcu_persistflags ; do 
-	cp ${MONITOR_DATA_ROOT}/production.$f.pkl ${MONITOR_ARCHIVE_ROOT}/`date +%F-%H:%M`.production.$f.pkl
-done
+# 1. FINDBAD NODES 
+${MONITOR_SCRIPT_ROOT}/commands/findall.py $DATE || :
+ps ax | grep BatchMode | grep -v grep | awk '{print $1}' | xargs -r kill || :
+# clean up stray 'locfg' processes that hang around inappropriately...
+ps ax | grep locfg | grep -v grep | awk '{print $1}' | xargs -r kill || :
 
-echo "Running grouprins on all dbg nodes"
-############################
-# 5. Check if there are any nodes in dbg state.  Clean up afterward.
-${MONITOR_SCRIPT_ROOT}/grouprins.py --mail=1 \
-	--nodeselect 'state=DEBUG&&boot_state=(rins|dbg|boot)' \
-	--stopselect 'state=BOOT&&kernel=2.6.22.19-vs2.3.0.34.9.planetlab' \
-	--reboot || :
-${MONITOR_SCRIPT_ROOT}/findbad.py --increment --cachenodes --debug=0 --dbname="findbad" --nodeselect 'state=DEBUG&&boot_state=dbg||state=DEBUG&&boot_state=boot' || :
 
-echo "Collecting RT database dump"
-##########################
-# 6. cache the RT db locally.
-python ${MONITOR_SCRIPT_ROOT}/rt.py
+${MONITOR_SCRIPT_ROOT}/commands/policy.py $DATE || :
+curl -s 'http://summer.cs.princeton.edu/status/tabulator.cgi?table=table_nodeview&formatcsv' > /var/lib/monitor/comon/$DATE.comon.csv || :
 
+cp ${MONITOR_SCRIPT_ROOT}/monitor.log ${MONITOR_ARCHIVE_ROOT}/`date +%F-%H:%M`.monitor.log
+service plc restart monitor || :
 rm -f $MONITOR_PID