|Intro | Install | Using | Monitor | Tuning | Advanced | Platform Notes

Teambuilder

[Prev: Using Teambuilder] [Home] [Next: Tuning Teambuilder]

Teambuilder Monitor

The Teambuilder Monitor (tbmonitor) displays the activity and statistics of the machines participating in the compilation farm. It also allows some daemon parameters to be adjusted.

All the parameters are set in configuration files, but sometimes it is convenient to make small or temporary adjustments, and the Teambuilder Monitor provides this functionality.

When tbmonitor is started it will broadcast a request for schedulers. The first scheduler found will be displayed in the monitor main window (Figure 1).

Figure 1. Monitor main window

If you have other schedulers in your network you can select them via View|Select scheduler... (Figure 2).

Figure 2. Scheduler selector

The monitor displays the state of the teambuilder network. Each participant in the network is represented with a different color. The graph displays:

• the machines currently producing jobs (bold hostname);

• compilation jobs (solid color);

• local jobs, e.g. generating jobs, linking (cross-hatched color).

• Failed daemons (not shown). The error message can be seen by clicking on the daemon. The only failure currently reported is "No space left on device". Clear some space in the tmp directory and restart the daemon.

Figure 3 shows the machine called drumstick, in blue, generating jobs which are being run on all machines in the farm. The maximum jobs allowed on any of the machines in the farm is three (most are dual processor). Aurochs and brahman have only one processor so they have MaxJobs set to 1 (dotted line). Some slower dual processor machines have MaxJobs set to 2. Glacier has maxJobs set to zero so that it does not participate in compilation (gray hostname).

Figure 3. Drumstick's jobs being compiled by the farm

This information is usually enough to see what is happening to your compilation. You can see more information on the farm by selecting View|Show statistics (Figure 4). This figure also show two machines producing jobs (drumstick and billabong).

Figure 4. Statistics

The statistics shown are:

• Up -- the time that the daemon has been running.

• Rcvd -- the number of bytes the daemon has received (for compilation).

• Sent -- the number of bytes the daemon has sent (object files).

• Jobs -- the number of jobs the daemon has processed.

• ms/job -- the average number of milliseconds taken for each job.

• B/ms -- a measure of the number of bytes processed per millisecond. This gives a rough indication of the speed of the machine.

• Load -- the load average of the machine.

Clicking on one of the daemons displays the daemon configuration dialog. This allows basic daemon properties to be modified. Figure 5 shows the daemon configuration for tundra. This dialog also displays the compilers that are available on the daemon and what files are currently being compiled. As you can see in the screenshot, tundra has two different versions of the ARM Linux cross-compiler installed, as well as a native compiler.

Figure 5. Daemon Configuration

Command Line Mode

The monitor can be used to adjust the basic daemon settings via the command line. The monitor command line mode useage is:

tbmonitor [-settings] [-joblimit l] [-priority p] daemon

• -settings Prints current settings.

• -joblimit l Set the number of jobs to run on the daemon.

• -priority p Set the priority of jobs to run on the daemon (0 - high, 19 - low).

This is useful when a machine will be busy for some time and running Teambuilder jobs is unwanted. For example, a script that performs a long CPU intensive operation may look like:

#!/bin/sh
tbmonitor -joblimit 0 localhost
# run intensive operation
tbmonitor -joblimit 1 localhost

[Prev: Using Teambuilder] [Home] [Next: Tuning Teambuilder]