Validate your hardware and network performance.
Greenplum provides a management utility called gpcheckperf, which can be used to identify hardware and system-level issues on the machines in your Greenplum Database array. gpcheckperf starts a session on the specified hosts and runs the following performance tests:
gpcheckperf, you must have a trusted host setup between the hosts involved in the performance test. You can use the utility gpssh-exkeys to update the known host files and exchange public keys between hosts if you have not done so already. Note that
gpcheckperf calls to gpssh and gpscp, so these Greenplum utilities must be in your
Parent topic: Installing and Upgrading Greenplum
To test network performance, run gpcheckperf with one of the network test run options: parallel pair test (
-r N), serial pair test (
-r n), or full matrix test (
-r M). The utility runs a network benchmark program that transfers a 5 second stream of data from the current host to each remote host included in the test. By default, the data is transferred in parallel to each remote host and the minimum, maximum, average and median network transfer rates are reported in megabytes (MB) per second. If the summary transfer rate is slower than expected (less than 100 MB/s), you can run the network test serially using the
-r n option to obtain per-host results. To run a full-matrix bandwidth test, you can specify
-r M which will cause every host to send and receive data from every other host specified. This test is best used to validate if the switch fabric can tolerate a full-matrix workload.
Most systems in a Greenplum Database array are configured with multiple network interface cards (NICs), each NIC on its own subnet. When testing network performance, it is important to test each subnet individually. For example, considering the following network configuration of two NICs per host:
|Greenplum Host||Subnet1 NICs||Subnet2 NICs|
You would create four distinct host files for use with the gpcheckperf network test:
You would then run gpcheckperf once per subnet. For example (if testing an even number of hosts, run in parallel pairs test mode):
$ gpcheckperf -f hostfile_gpchecknet_ic1 -r N -d /tmp > subnet1.out $ gpcheckperf -f hostfile_gpchecknet_ic2 -r N -d /tmp > subnet2.out
If you have an odd number of hosts to test, you can run in serial test mode (
Parent topic: Validating Your Systems
To test disk and memory bandwidth performance, run gpcheckperf with the disk and stream test run options (
-r ds). The disk test uses the
dd command (a standard UNIX utility) to test the sequential throughput performance of a logical disk or file system. The memory test uses the STREAM benchmark program to measure sustainable memory bandwidth. Results are reported in MB per second (MB/s).
Log in on the master host as the
greenplum_path.sh path file from your Greenplum installation. For example:
$ source /usr/local/greenplum-db/greenplum_path.sh
Create a host file named
hostfile_gpcheckperf that has one host name per segment host. Do not include the master host. For example:
sdw1 sdw2 sdw3 sdw4
gpcheckperf utility using the
hostfile_gpcheckperf file you just created. Use the
-d option to specify the file systems you want to test on each host (you must have write access to these directories). You will want to test all primary and mirror segment data directory locations. For example:
$ gpcheckperf -f hostfile_gpcheckperf -r ds -D \ -d /data1/primary -d /data2/primary \ -d /data1/mirror -d /data2/mirror
The utility may take a while to perform the tests as it is copying very large files between the hosts. When it is finished you will see the summary results for the Disk Write, Disk Read, and Stream tests.
Parent topic: Validating Your Systems