Ganglia — Accessing Raw Data

Ganglia is a scalable distributed system monitor tool for high-performance computing systems such as clusters and grids. It enables the user to remotely view live or historical statistics (such as CPU load averages or network utilization) for all machines that are being monitored.

Time series metrics from Ganglia are stored in a round-robin database (RRD) using the RRDtool. The SAS IT Resource Management adapter for Ganglia reads the RRD that was created using Ganglia.

The Ganglia RRD data can be gathered at any interval (step), any metrics can be gathered, and any consolidation function (CF) can be used. Because SAS IT Resource Management has its own aggregation process, the data read from the round-robin databases should be detail data, not consolidated data. To store detail data using the RRDtool, the RRDs should store the data with the CF set to Average. In addition, the average should be generated based on one step of data, where a step is the interval of time that was specified when the RRD was generated. If the data in the RRD is consolidated, then SAS IT Resource Management requires that the CF staging parameter be set accordingly for that consolidation. If this staging parameter is left blank, then the data in all of the CFs in the RRD are collected.

The adapter can read a single round-robin database, or it can read all round-robin databases in a directory. If multiple round-robin databases are read, the data is combined into a single staging table.

Because a round-robin database can store only numeric data, users of the RRDtool generally store identifying character data in the name or path of the round-robin database. The filename of each round-robin database is saved by the Ganglia adapter in a field called filename. The filename field that is saved by the Ganglia adapter creates Grid performance metrics or any relevant computed columns with this identifying information.

To read the raw data from the RRDtool database, Perl scripts are run as part of the staging process. This means that Perl must be installed on the machine that the SAS program runs on (that is, the SAS Workspace Server, the SAS batch server, or interactive SAS).

The Perl scripts that are used by the staging code are generated as part of the code generation of the staging job. This code can be viewed and modified as necessary in the Code tab of the job.

To run these scripts in batch mode, set the XCMD option for the Batch server. To do so, change set USERMODS_OPTIONS= to set USERMODS_OPTIONS=XCMD.

To run these scripts from within the SAS Workspace Server, the SAS option XCMD must be turned on. By default, this option is off in the SAS Workspace Server. For instructions on how to turn on the XCMD option, see How to Turn On the XCMD Option.

The SSH host command is an executable object that is available as part of the functionality of the Ganglia, RRDtool, and SNMP adapters. This command specifies the RSH or SSH version of the command and the name of the host for running the rrdtool command. The rrdtool command facilitates reading data from round-robin database files that are located on other hostnames. Entering the SSH version of the command triggers this SSH functionality. This is the format of the command: ssh user@hostname. It is entered in the rsh/ssh host command field on the Staging Parameters tab of the Properties dialog box for the adapter’s staging transformation.

If you enter a value in the rsh/ssh host command field that begins with SSH, then SAS IT Resource Management assumes that this job is running on a UNIX environment. The Perl script is changed so that it uses the UNIX find command to get the list of round-robin database files from the other hostnames.

To enable this functionality, set up SSH authentication using the SSH key-gen file. Then copy the file that stores the key to the host where the RRDtool executable and raw data are located. After the SSH key-gen file is copied, make sure that you can access the target host that is issuing the SSH command from the source host. If you have alias names for your target host, make sure that you execute the SSH command manually from the source host with that alias. By doing this, the command generates the host/RSA key for that alias hostname. Thus, the expected warning or error message is avoided during execution of the staging job. The following message is an example of the warning or error message:

Host key verification failed

To use this SSH method, go to the Staging Parameters tab of the Properties dialog box for the adapter’s staging transformation. Then fill in values for the following options: