github
/
prometheus
mirror of https://github.com/prometheus/prometheus
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
Browse Source

update the doc link in internal_arthitecture.md (#7966) 

* update the doc link in internal_arthitecture.md
* address reviewer's comment to remove out-dated wrapper

Signed-off-by: Luke Chen <showuon@gmail.com>
pull/7934/head
Luke Chen 4 years ago committed by GitHub
parent
7e2db3d092
commit
3364875ae5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 5 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 10
      documentation/internal_architecture.md

10
documentation/internal_architecture.md
Unescape View File

@ -40,7 +40,7 @@ The [termination handler](https://github.com/prometheus/prometheus/blob/v2.3.1/c
The scrape discovery manager is a [`discovery.Manager`](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/manager.go#L73-L89) that uses Prometheus's service discovery functionality to find and continuously update the list of targets from which Prometheus should scrape metrics. It runs independently of the scrape manager (which performs the actual target scrapes) and feeds it with a stream of [target group](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/targetgroup/targetgroup.go#L24-L33) updates over a [synchronization channel](https://github.com/prometheus/prometheus/blob/v2.3.1/cmd/prometheus/main.go#L431).
Internally, the scrape discovery manager runs an instance of each configuration-defined service discovery mechanism in its own goroutine. For example, if a `scrape_config` in the configuration file defines two [`kubernetes_sd_config` sections](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#%3Ckubernetes_sd_config%3E), the manager will run two separate [`kubernetes.Discovery`](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/kubernetes/kubernetes.go#L150-L159) instances. Each of these discovery instances implements the [`discovery.Discoverer` interface](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/manager.go#L41-L55) and sends target updates over a synchronization channel to the controlling discovery manager, which then enriches the target group update with information about the specific discovery instance and forwards it to the scrape manager.
Internally, the scrape discovery manager runs an instance of each configuration-defined service discovery mechanism in its own goroutine. For example, if a `scrape_config` in the configuration file defines two [`kubernetes_sd_config` sections](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#kubernetes_sd_config), the manager will run two separate [`kubernetes.Discovery`](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/kubernetes/kubernetes.go#L150-L159) instances. Each of these discovery instances implements the [`discovery.Discoverer` interface](https://github.com/prometheus/prometheus/blob/v2.3.1/discovery/manager.go#L41-L55) and sends target updates over a synchronization channel to the controlling discovery manager, which then enriches the target group update with information about the specific discovery instance and forwards it to the scrape manager.
When a configuration change is applied, the discovery manager stops all currently running discovery mechanisms and restarts new ones as defined in the new configuration file.
@ -71,7 +71,7 @@ In the same way that the scrape discovery manager runs one discovery mechanism f
### Target labels and target relabeling
Whenever the scrape manager receives an updated list of targets for a given scrape pool from the discovery manager, the scrape pool applies default target labels (such as `job` and `instance`) to each target and applies [target relabeling configurations](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#%3Crelabel_config%3E) to produce the final list of targets to be scraped.
Whenever the scrape manager receives an updated list of targets for a given scrape pool from the discovery manager, the scrape pool applies default target labels (such as `job` and `instance`) to each target and applies [target relabeling configurations](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config) to produce the final list of targets to be scraped.
### Target hashing and scrape timing
@ -93,15 +93,15 @@ Currently rules still read and write directly from/to the fanout storage, but th
### Local storage
Prometheus's local on-disk time series database is a [light-weight wrapper](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/tsdb/tsdb.go#L102-L106) around [`github.com/prometheus/prometheus/tsdb.DB`](https://github.com/prometheus/prometheus/blob/master/tsdb/db.go#L92-L117). The wrapper makes only minor interface adjustments for use of the TSDB in the context of the Prometheus server and implements the [`storage.Storage` interface](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/interface.go#L31-L44). You can find more details about the TSDB's on-disk layout in the [local storage documentation](https://prometheus.io/docs/prometheus/latest/storage/).
About Prometheus's local on-disk time series database, please refer to [`github.com/prometheus/prometheus/tsdb.DB`](https://github.com/prometheus/prometheus/blob/master/tsdb/db.go). You can find more details about the TSDB's on-disk layout in the [local storage documentation](https://prometheus.io/docs/prometheus/latest/storage/).
### Remote storage
The remote storage is a [`remote.Storage`](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/remote/storage.go#L31-L44) that implements the [`storage.Storage` interface](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/interface.go#L31-L44) and is responsible for interfacing with remote read and write endpoints.
For each [`remote_write`](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#%3Cremote_write%3E) section in the configuration file, the remote storage creates and runs one [`remote.QueueManager`](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/remote/queue_manager.go#L141-L161), which in turn queues and sends samples to a specific remote write endpoint. Each queue manager parallelizes writes to the remote endpoint by running a dynamic number of shards based on current and past load observations. When a configuration reload is applied, all remote storage queues are shut down and new ones are created.
For each [`remote_write`](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write) section in the configuration file, the remote storage creates and runs one [`remote.QueueManager`](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/remote/queue_manager.go#L141-L161), which in turn queues and sends samples to a specific remote write endpoint. Each queue manager parallelizes writes to the remote endpoint by running a dynamic number of shards based on current and past load observations. When a configuration reload is applied, all remote storage queues are shut down and new ones are created.
For each [`remote_read`](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#%3Cremote_read%3E) section in the configuration file, the remote storage creates a [reader client](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/remote/storage.go#L96-L118) and results from each remote source are merged.
For each [`remote_read`](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_read) section in the configuration file, the remote storage creates a [reader client](https://github.com/prometheus/prometheus/blob/v2.3.1/storage/remote/storage.go#L96-L118) and results from each remote source are merged.
## PromQL engine

Write Preview
Loading…
Cancel
Save