Gratia Transfer Monitoring for Hadoop
Notes on how to get Gratia Transfer Monitoring of the Hadoop GridFtp servers
The grid-enabled Hadoop DFS utilizes normal Globus GridFTP servers with a specialized plugin in order to do transfers. Because it uses this server software, we are able to take advantage of the OSG's accounting system. The system, Gratia, has a plugin which monitors the Globus GridFTP logs and reports any transfer data to a central collector.
The instructions on this page must be followed for every GridFTP server on Hadoop.
The instructions on this page must be followed for every GridFTP server on Hadoop.
- Source the OSG install:
source /opt/osg/osg-100/setup.sh
- Configure the probe with the newly installed configuration script.
$VDT_LOCATION/vdt/setup/configure_gratia --probe gridftp-transfer --report-to <gratia_host:gratia_port> --probe-cron --site-name "blah"
For Nebraska, this should be:$VDT_LOCATION/vdt/setup/configure_gratia --probe gridftp-transfer --report-to rcf-gratia.unl.edu:8880 --probe-cron --site-name "Nebraska"
- Check to make sure the file
$VDT_LOCATION/gums/config/gums-client.properties
has the correct GUMS server. If it does not, contact Brian: the VDT_GUMS_HOST setting when you installed the GridFTP_HDFS package should have taken care of this. - Turn the services on:
vdt-control --on gratia-gridftp-transfer
vdt-control --enable gums-host-cron
vdt-control --on gums-host-cron - This should install a cronjob for the user root. If you run the following command as root:
crontab -l
The output should include a line like this:7,17,27,37,47,57 * * * * /opt/osg/osg-100/gratia/probe/gridftp-transfer/gridftp-transfer_meter.cron.sh > /opt/osg/osg-100/gratia/var/logs/gratia-probe-gridftp-transfer.log 2>&1
36 6,12,18,0 * * * /opt/osg/osg-100/gums/scripts/gums-host-cron - Run gums-host-cron by hand:
/opt/osg/osg-100/gums/scripts/gums-host-cron
It has no output, but you should verify that the generated file is non-zero:ls -l $VDT_LOCATION/monitoring/osg-user-vo-map.txt
Also, read through the logfile $VDT_LOCATION/monitoring/log/gums-host-cron.log to make sure there were no errors. - Run the Gratia probe once by hand to check for functionality:
/opt/osg/osg-100/gratia/probe/gridftp-transfer/gridftp-transfer_meter.cron.sh
Look for any abnormal termination and report it if it is a non-trivial site issue. - (Nebraska admins only) Check the following page to make sure the new GridFTP server probe shows up on the list of distinct probe names. It will have the name gridftp-transfer:example.com, where example.com is replaced by the host's FQDN:
http://rcf-gratia.unl.edu:8880/gratia-reporting/showQuery.jsp?sql=select%20distinct%20ProbeName%20from%20JobUsageRecord_Meta;&ReportTitle=Distinct%20Probe%20Names&displayReport=false&__title
Quick-Install script for Nebraska
source /opt/osg/osg-100/setup.sh
cd $VDT_LOCATION
$VDT_LOCATION/vdt/setup/configure_gratia --probe gridftp-transfer --report-to rcf-gratia.unl.edu:8880 --probe-cron --site-name "Nebraska"
vdt-control --on gratia-gridftp-transfer
vdt-control --enable gums-host-cron
vdt-control --on gums-host-cron
$VDT_LOCATION/gums/scripts/gums-host-cron