HostScope

HostScope is a tool to gather and display key system statistics.

The performance information is made available on the network via multi-casting. I.e. a hostscope client can visualize the system state of multiple hosts on the network that run the hostscoped server.

Hostscope provides a large amount of system information and statistics that allow a detailed analysis of system performance and bottlenecks that may limit maximum performance. Examples of merics of hostscope are: - Operating system metrics such as forks, context switches and interrups - CPU load, usage (e.g. user, system, interrupts), and temperature - Memory usage and paging performance - Power supply states and battery charging levels - Disk and Network I/O performance - Per user summary information regarding CPU, I/O, memory, and faults

For a detailed description of all metrics, please refer to the help and manual pages of hostscope.

Influx integration:

HostScope provides a bridge executable hostscopebr for submitting data to an Influx DB. Using Grafana the data provided by hostscope can be visualized in a full custom way.

The bridge utility should be run on the same machine that runs the Influx DB server, but can also be run on another machine. It gathers data from all hostscope daemons - i.e. only one instance per network (multi-cast domain) is necessary.

Influx must be configured to receive data via UDP. For this, search for the [[udp]] section in /etc/influxdb/influx.conf and set a network port and the database to use. The recommended precision for the data collection is milliseconds. After restarting Influx DB start hostscopebr with option -I <host>:<port> to specify where Influx is listening. The argument <host> can be omitted if the bridge is running on the same machine as Influx.

Hostscope submits its measurements with the name "hostscope" to Influx. The name of the database is chosen when setting up the UDP port in influx configuration. Every measurement is tagged with the associated hostname with a tag named "host".

Performance Analysis:

For best system performance you should make sure that your system has no bottlenecks that limit its maximum performance. Hostscope will give you quick insight into system performance by providing detailed information on the key performance contributors: CPU, memory, and I/O.

For maximum performance neither of these three should be limited. For CPU a certain amount of idle should be visible. For memory make sure that no paging activity is visible. For I/O make sure that the storage devices are not 100% busy and that the network devices don't drop any packets.

Following Influx measurements are provided:

• uptime: uptime in seconds
• lavg1m: load average over past minute
• lavg5m: load average over past 5 minutes
• lavg15m: load average over past 15 minutes
• memfree: amount of free RAM in bytes
• pagedout: number of bytes pushed to swap space
• fork: process forks per second
• ctxtsw: context switches per second
• pgpgin: pages paged in per second
• pgpgout: pages paged out per second
• pgfree: number of pages freed per second
• pgfault: number of page faults per second
• intr: number of hardware interrupts processed per second
• sirqs: number of software interrupts processed per second
• cpu_user: percentage of user time over all CPUs
• cpu_sys: percentage of system time over all CPUs
• cpu_irq: percentage of time processing interrupts over all CPUs
• cpu_idle: percentage of idle time all CPUs
• net_rxps: total B/s received on all network interfaces
• net_txps: total B/s transmitted on all network interfaces
• disk_rdps: total B/s read from all storage devices
• disk_wrps: total B/s written to all storage devices

Status:

• This software is stable and its binary protocol is committed.
• Influx metrics are evolving.

Feedback:

Feedback concerning this software is welcome. Please send an e-mail to thomas@maier-komor.de if you have any comments, feedback, ideas or if you should encounter a bug.