http://www.x-trace.net/wiki/ 2008-10-03T07:51:15-07:00 http://www.x-trace.net/wiki/ http://www.x-trace.net/wiki/lib/images/favicon.ico text/html 2008-10-02T16:37:40-07:00 andy http://www.x-trace.net/wiki/doku.php?id=start&rev=1222990660&do=diff About X-Trace is a network diagnostic tool designed to provide users and network operators with better visibility into increasingly complex Internet applications. It does this by annotating network requests with metadata that can be used to reconstruct requests, even those that make use of multiple network layers. X-Trace “enabled” Internet sites make use of these identifiers to record the path that requests take through their network. text/html 2008-09-30T19:05:56-07:00 andy http://www.x-trace.net/wiki/doku.php?id=docs:tutorial-1&rev=1222826756&do=diff X-Trace Java API (javadoc) text/html 2008-03-18T13:08:37-07:00 George Porter http://www.x-trace.net/wiki/doku.php?id=docs:server-config-guide-2.0&rev=1205870917&do=diff X-Trace 2.0 Server configuration guide Configuring the X-Trace stand-alone backend server is done entirely through command-line options. These options are specified as Java parameters on the command line when the server is first started. At the moment, you cannot update these settings at runtime. text/html 2008-02-21T09:50:39-07:00 George Porter http://www.x-trace.net/wiki/doku.php?id=disc:release2roadmap&rev=1203616239&do=diff * Ensure that all source files have a license at the top * Update the license to include '2008' in the copyright area of the license * Check that the Graphviz Java applet's license is ok for distribution * For all sublibraries and components, put their license in LICENSE.Contributed * Update and fill in the HTML template for the embedded webserver * Put in the x-trace logo * Add some search boxes on the 'home page' to search by task id, tag, title, etc. * Add pagination suppo… text/html 2008-02-15T20:54:35-07:00 George Porter http://www.x-trace.net/wiki/doku.php?id=disc:start&rev=1203137675&do=diff * OpId size * High Performance Reporting * Ideas for release 2 * sandbox * Roadmap to release 2.0 text/html 2008-01-12T13:55:56-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=docs:report_specifications&rev=1200174956&do=diff X-Trace Report Key/Value Schema X-Trace reports consist of key/value pairs. The purpose of this page is to capture de-facto schema for these pairs. Each of the cases below is a proposal for: * Reserving the use of a given key for some purpose * Describing the contents of that key * Describing the data format for the value of the key * Providing for valid and invalid values * Indicating whether the key can appear multiple times in a given report * Outlining what should be done w… text/html 2007-12-16T16:18:55-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=disc:sandbox&rev=1197850735&do=diff Part 1: Part 1 * Unordered List Item * Unordered List Item * Another one. * FoobarA link to sadbox text/html 2007-12-05T18:59:28-07:00 George Porter http://www.x-trace.net/wiki/doku.php?id=disc:release2ideas&rev=1196909968&do=diff X-trace notes * getting rid of Xtr: not for generic stuff like Context or Event * Xtrace(Context?) class with various inner classes (Process, Event) * Process.end() instead of passsing the Process object back to the XtraceContext * A version of logEvent() that takes an array of extension fields and edges in one operation. * Check on the performance of XtrMetadata.toString() * Check if the hostname is cached * Make the opId a byte array of either byte[4] or byte[8] or use an intege… text/html 2007-12-01T19:50:56-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=docs:severity&rev=1196567456&do=diff This is inspired by syslog and Log4J priorities and levels. However, the semantics are slightly different, as the level here is not always associated with the code point that generates the report, but can be associated with the metadata. It is common in logging systems to define different levels of importance or detail for logging messages, and to have some way of specifying, either at compile time or run time, which levels should be logged. Too much logging imposes too much overhead on the sy… text/html 2007-12-01T14:51:50-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=disc:high-performance_reporting&rev=1196549510&do=diff The communication between the application and the X-Trace frontend daemon is currently over an UDP socket. The problem is that each report entails a context switch and some copying to the kernel. Some other options are: Local TCP socket Advantages: text/html 2007-11-21T09:38:13-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=docs:logging-api&rev=1195666693&do=diff Instrumenting your code with X-Trace This page introduces the X-Trace logging-like API. We use the Java API as an example, but the C++ API is similar. Introduction The X-Trace logging API makes instrumenting code with X-Trace as easy as using a logging API in the common case, while allowing you to capture complex annotation or execution structures with additional calls. text/html 2007-11-20T18:30:15-07:00 Rodrigo Fonseca http://www.x-trace.net/wiki/doku.php?id=disc:opid_size&rev=1195612215&do=diff OpIds are the mechanism used by X-Trace to identify an event within a task, and to link them together. Originally they were drawn from a 32-bit namespace, but that generated collisions as some tasks started to have ~ 10^5 events. The plot below shows the probability of at least one collision -- two events with the same OpId. 'Number of Draws' is the number of events in a task. The conclusion is that 4-byte OpIds are only suitable if the task sizes expected in the system in consideration are sm…