loghisto
loghisto copied to clipboard
spacejam
counters and logarithmically bucketed histograms for distributed systems
Currently, a ring buffer of a fixed length is used for the outbound queue. Metrics are dropped when this is filled. Possibly just use https://github.com/bitly/nsq/blob/master/nsqd/diskqueue.go
This is relatively low-effort and potentially a highly desired feature. The current default is a good general choice, although the 0-1 precision degradation may rub users who do not primarily...
Currently the external stats senders for Graphite and OpenTSDB/KairosDB just attempt to slam their buffers into a socket, and if the local OS doesn't have an issue with that, it...