TReqS » TReqS-legacy

A temporary patch was created. It checks the databases to see the old file requests and then if there are unprocessed file requests the script will restart TReqS.
This is not a very clean strategy, but it is fast.

The file is called treqs_joker.sh and it is found in /opt/treqs/etc

email=hpssmaster@cc.in2p3.fr
min=30

res=`echo "select count(1) from requests where creation_time < date_sub(current_timestamp(), interval $min minute) and status = 10" | mysql --host=cctreqs --user=treqsweb --database=treqsjobs | awk '!/count/ {print}'`
if ${res} != "0"
then
echo "TReqS was restarted because there were $res requests not processed in the last $min minutes" | mail -s "TReqS restarted due to problematic behaviour" $email
/etc/init.d/treqsd restart
fi

The new jTReqS version does not have this problem. The references are stored in a MultiMap and once they are deleted, they are completly deleted.
The current version in C++ still keeps the problem.