On the last Golang Barcelona meetup we decided to try new format of conversation and after the talk started open discussion on logging, tracing and metrics. The idea was to encourage people to read a very nice blog post by Peter Bourgon on this subject - "Metrics, Tracing and Logging" and discuss it in a free format.
Go dependency tool
This document proposes a change to memory allocation to eliminate the need for the sweeper and a new representation for the mark bitmap that enables this. This reduces the cost of allocation, significantly improves the locality of the mark bitmap, and paves the way for future advances to the Go GC that require multiple mark bits.
package main import ( "encoding/json" "io" "io/ioutil" "log" "net/http" ) type msg struct { Text string } func handleErr(err error) { if err != nil { log.Fatalf("%s\n", err) } } // use a io.Pipe to connect a JSON encoder to an HTTP POST: this way you do // not need
Go's empty interface{} is the interface that everything implements. It allows functions that can be passed any type. The function func f(any interface{}) can be called with a string f("a string"), an integer f(42), a custom type, or anything else. This flexibility comes at a cost.
Dmitry Vyukov, dvyukov@ Go has a number of profiling tools -- CPU, memory, blocking profilers; GC and scheduler tracers and heap dumper. However, profilers provide only aggregate information, for example, how much memory in total was allocated at this source line. Tracers provide very shallow information.
Published on - Filed under golang, bare metal This post is a post-mortem of a Go production system crash. We'll use the logs to drill down into the guts of Go's network source code to find a leak that affected a core part of our stack at Timehop.