Using the Workload Manager Graphical Interface (gptop)
The Workload Manager Graphical interface,
gptop, is a curses interface that you can use to monitor live data for the rules engine, host statistics, active Greenplum Database queries, and database skew.
You can start
gptop from the command line by running
gptop in a terminal. If you are already using interactive
gp-wlm, enter the
gptop command to enter the monitor.
Note: If you use the PuTTY ssh/telnet client for Windows to run
gptop, you may experience problems with function keys and line-drawing characters with the default settings. To support function keys, in the PuTTY Configuration window, choose Connection > Data and enter
putty in the Terminal-type string field. To enable correct line-drawing characters, choose Window > Translation and set Remote character set to Use font encoding.
When you first start
gptop, the GPDB Queries pane (see below) is selected. At any time, you can press the F2 key to get a pane selection menu. Use the Tab, Left-Arrow, or Right-Arrow keys to make a selection. Press F2 to close an open menu without making a selection.
An asterisk (
*) next to a column heading indicates that the rows are sorted by that column. To change the sort order, press the F3 key, then choose the number of the column you want to sort by from the pop-up menu.
q or choose File > Exit to leave
gptop monitoring features are under the Monitor menu. The Monitor menu has four options:
- GPDB Queries – Shows active Greenplum Database queries
- GPDB Skew – Shows skew statics for active queries
- Hydra – Shows statistics from the rules engine
- SysData – Shows performance statistics for each host in the cluster
Note: Queries that run in under five seconds are not reported by
gptop in order to minimize load on the system and to focus on queries consuming greater resources.
The GPDB Queries monitor displays a line for each active Greenplum Database query.
The GPDB Skew monitor shows calculated skew statistics for active Greenplum Database queries. Statistics are calculated on each host in the system and then sent to the master where they are summarized. You can select a host and press Enter to see statistics for the host. The calculated skew value is the cubed standard deviation across the cluster. Values closer to 0.0 indicate less skew. The GPDB Skew monitor shows the following columns for each active query: