github.com/dropbox/llama
Llama client to pull metrics from Llama collectors LLAMA Collector sends UDP probes to a set of target reflectors and provides statistics about their latency and reachability via an API. Mock Llama client used in `client_test.go` Functionality for sending and receiving UDP probes on a socket. portgroup defines PortGroup, which is used to multiplex UDPAddr structs to multiple ports via parallel channels. LLAMA Scraper pulls stats from Collectors and then writes them to the indicated database. Tags is a helper description for a structure that stores a map of attributes and values for a given key. Example: Tags["1.2.3.4"]["dst_hostname"] = "localhost" Tags["1.2.3.4"]["dst_cluster"] = "mycluster"
Readme
LLAMA (Loss and LAtency MAtrix) is a library for testing and measuring network loss and latency between distributed endpoints.
It does this by sending UDP datagrams/probes from collectors to reflectors and measuring how long it takes for them to return, if they return at all. UDP is used to provide ECMP hashing over multiple paths (a win over ICMP) without the need for setup/teardown and per-packet granularity (a win over TCP).
Black box testing is critical to the successful monitoring and operation of a network. While collection of metrics from network devices can provide greater detail regarding known issues, they don't always provide a complete picture and can provide an overwhelming number of metrics. Black box testing with LLAMA doesn't care how the network is structured, only if it's working. This data can be used for building KPIs, observing big-picture issues, and guiding investigations into issues with unknown causes by quantifying which flows are/aren't working.
At Dropbox, we've found this useful on multiple occasions for gauging the impact of network issues on internal traffic, identifying the scope of impact, and locating issues for which we had no other metrics (internal hardware failures, circuit degradations, etc).
Even if you operate entirely in the cloud LLAMA can help identify reachability and network health issues between and within regions/zones.
If you're looking to get started quickly with a basic setup that doesn't involve special integrations or customization, this should get you going. This assumes you have a running InfluxDB instance on locahost listening on port 5086 with a llama database already created.
In your Go development environment, in separate windows:
go run github.com/dropbox/llama/cmd/reflectorgo run github.com/dropbox/llama/cmd/collectorgo run github.com/dropbox/llama/cmd/scraperIf you want to run each of these on a separate machine/instance, after distributing the binaries created with go build, customizing the flags as needed:
reflector -port <port> to start the reflector listening on a non-default port.collector -llama.dst-port <port> -llama.config <config> where the port matches what the reflector is listening on, and the config is a YAML configuration based on one of the examples under configs/.scraper -llama.collector-hosts <hosts> -llama.collector-port <port> -llama.influxdb-host <hostname> -llama.influxdb-name <db-name> -llama.influxdb-pass <pass> -llama.influxdb-port <port> -llama.influxdb-user <user> -llama.interval <seconds>
collector-hosts being a comma-separated list of IP addresses or hostnames where collectors can be reachedcollector-port identifying the port on which the collector's API is configured to listeninfluxdb-* detailing where the InfluxDB instance can be reached, credentials, and databaseinterval being how often, in seconds, the scraper should pull data from collectors and write to the database. Should align with the summarization interval in the collector config.LLAMA was primarily built during a Dropbox Hack Week and is still considered unstable, as the API, config format, and overall design is not considered final. It works and we've been using the original internal version for quite a while, but we want to make various changes and improvements before considering a v1.0.0 release.
At this time, we're not ready for external contributors. Once we have a v1.0.0 release, we'll happily reconsider this and update accordingly. When that happens, substantial contributors will need to agree to the Dropbox Contributor License Agreement.
FAQs
Llama client to pull metrics from Llama collectors LLAMA Collector sends UDP probes to a set of target reflectors and provides statistics about their latency and reachability via an API. Mock Llama client used in `client_test.go` Functionality for sending and receiving UDP probes on a socket. portgroup defines PortGroup, which is used to multiplex UDPAddr structs to multiple ports via parallel channels. LLAMA Scraper pulls stats from Collectors and then writes them to the indicated database. Tags is a helper description for a structure that stores a map of attributes and values for a given key. Example: Tags["1.2.3.4"]["dst_hostname"] = "localhost" Tags["1.2.3.4"]["dst_cluster"] = "mycluster"
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
A hospital in Mobile, Alabama, agreed to a settlement in a landmark ransomware death lawsuit, but is now reportedly reconsidering the agreement and refusing to pay.
Security News
A new report explores how advancements in LLMs are enhancing cyber threats, including polymorphic malware, personalized spearphishing, and the risk of hijacking customer service bots.
Security News
ESLint has approved an RFC that adds support for TypeScript configuration files, which is aimed at improving the developer experience and recognizing changes in the evolving JavaScript ecosystem.