Index: head/en_US.ISO8859-1/htdocs/news/status/report-2016-10-2016-12.xml =================================================================== --- head/en_US.ISO8859-1/htdocs/news/status/report-2016-10-2016-12.xml (revision 49806) +++ head/en_US.ISO8859-1/htdocs/news/status/report-2016-10-2016-12.xml (revision 49807) @@ -1,154 +1,231 @@ October-December 2016
Introduction

This is a draft of the October–December 2016 status report. Please check back after it is finalized, and an announcement email is sent to the &os;-Announce mailing list.

This report covers &os;-related projects between October and December 2016. This is the last of four reports planned for 2016.

The last quarter of 2016 was another productive quarter for the &os; project and community. [...]

Thanks to all the reporters for the excellent work!

The deadline for submissions covering the period from January to March 2017 is April 7, 2017.

?>
team &os; Team Reports proj Projects kern Kernel arch Architectures bin Userland Programs ports Ports doc Documentation misc Miscellaneous &os;/EC2 Colin Percival cperciva@FreeBSD.org

This report covers work since the last &os;/EC2 status report (2015Q1).

&os;/EC2 is now part of the regular &os; release build, with snapshots and releases being automatically uploaded and copied to all available regions. Due to legal restrictions this does not currently include the GovCloud or China (Beijing) regions; anyone wishing to use &os; in those regions is encouraged to contact the author.

The AWS Marketplace reports that approximately 800 users are running roughly 2000 &os; EC2 instances. This does not count the likely significantly larger number of EC2 instances launched "directly" through the EC2 API and Console, but at least places a lower bound on usage.

&os; 11.0-RELEASE shipped with support for the "enhanced networking" support in EC2 C3, C4, R3, I2, D2, and M4 (excluding m4.16xlarge) instances; this provides significantly higher network performance than the virtual networking available on older EC2 instances and with older versions of &os;.

&os; 11.0-RELEASE and later also use indirect segment disk I/Os, which yield ~20% higher throughput with equal or lower latency, and support the 128-vCPU x1.32xlarge instance type.

&os; now supports the Amazon Simple Systems Manager service ("run command").

Complete a pending reorganization of the accounts used for &os;/EC2 releases. Support "second generation enhanced networking" via the new Elastic Network Adapter found in P2, R4, X1, and m4.16xlarge instances. Provide tools for improved functionality via the Simple Systems Manager service: Listing installed packages, checking for updates, adding/removing users, [your favourite sysadmin task goes here]. Add support for EC2's IPv6 networking to the default &os;/EC2 configuration. Continue ongoing interoperability testing between &os;'s NFS client and Amazon Elastic File System (NFS-as-a-service). + + + Sysctl Exporter for Prometheus + + + + + Ed + Schouten + + ed@FreeBSD.org + + + + + The Prometheus Project + Node Exporter + Sysctl Exporter + + + +

Prometheus is an Open Source monitoring system that was + originally built at SoundCloud in 2012. Since 2016, this project is + part of the Cloud Native Computing Foundation, together with other + projects like Kubernetes.

+ +

Prometheus scrapes its targets by periodically sending HTTP GET + requests. Targets then respond by sending key-value pairs of metrics + and their sample value. Prometheus has a query language, PromQL, + that can be used to aggregate sample values and specify alerting + conditions. Tools like Grafana can be used to create fancy + dashboards using such queries.

+ +

The Prometheus project provides a utility called the + node_exporter that gathers basic system metrics and serves them + over HTTP. This utility tends to be rather complex, as it has + to extract metrics from many different sources. On Linux, files + in /proc have no uniform format, meaning that for every + kernel framework a custom collector needs to be written.

+ +

On &os; the sitiuation is better, as the data exported + through sysctl is already structured in such a way that + it can easily be translated to Prometheus' metrics format. The + goal of this project is thus to provide a generic exporter for + the entire sysctl tree. Not only does this prevent unnecessary + bloat and indirection, it may also make the life of a kernel + developer a lot easier. One can easily use Prometheus to graph + the occurrence of an event over time by (temporarily) adding a + counter to the kernel.

+ +

An initial version of the sysctl exporter has been integrated + into the &os; base system in December. It can be run through + inetd by uncommenting the example provided in + inetd.conf. Unfortunately, this exporter cannot be + merged back to &os; 10.x/11.x, as it depends on KBI-breaking + changes to sysctl(9).

+ + + + Are you using Prometheus or are you interested in using + it? Be sure to give both Prometheus and this sysctl + exporter a try! + + It would be nice if we created a set of useful alerting + rules and placed those in /usr/share/examples. For example, how + can one use this exporter to monitor the state of GEOM-based + RAID arrays? Is such information even exported through + sysctl? + + Prometheus uses a pretty nifty format for exporting + histograms. Histograms are useful for expressing the amount of + time taken to complete certain events (for example, disk + operations). Would it be possible to add histograms as native + datatypes to sysctl? If so, is there any chance they can be + implemented without picking up any kernel locks? + +