add docs for loki deployment, especially for separate mode

docs/loki-deployment.md

@@ -0,0 +1,160 @@
+# Overview
+This page lists the 2 common Loki deployments, to help user to better understand Loki architecture.
+
+Loki consists several components as below (Quote from slides [here](https://grafana.com/blog/2018/12/12/loki-prometheus-inspired-open-source-logging-for-cloud-natives/)):
+![Loki Arch](loki_arch.png)
+
+|  Component | Intro |
+| --- | --- |
+| Distributor | use consistent hashing to assign a logstream to an ingester |
+| Ingester | gzip logs and append them. Index and data chunks go to different store |
+| Querier | look at the index to figure out which chunks match, and distributed grep through them. Talks to chunks and Ingester(get the recent data that has not been flushed) both |
+| Table-Manager | manage the chunk and index lifecycle |
+
+
+# Loki Deployment on Single Process
+To those who want to just have a try at Loki, deploying Loki on `Single Process` is the easiest way.
+It combines all above component in a single process.
+
+[Run Locally Using Docker](https://github.com/grafana/loki/tree/master/production#run-locally-using-docker) and [Run from Source](https://github.com/grafana/loki/tree/master/production#build-and-run-from-source) have already gave the examples.
+
+They are using Loki config like below
+```yaml
+auth_enabled: false
+
+server:
+  http_listen_port: 3100
+
+ingester:
+  lifecycler:
+    address: 127.0.0.1
+    ring:
+      kvstore:
+        store: inmemory
+      replication_factor: 1
+  chunk_idle_period: 15m
+
+schema_config:
+  configs:
+  - from: 2018-04-15
+    store: boltdb
+    object_store: filesystem
+    schema: v9
+    index:
+      prefix: index_
+      period: 168h
+
+storage_config:
+  boltdb:
+    directory: /tmp/loki/index
+
+  filesystem:
+    directory: /tmp/loki/chunks
+
+limits_config:
+  enforce_metric_name: false
+  reject_old_samples: true
+  reject_old_samples_max_age: 168h
+
+chunk_store_config:
+  max_look_back_period: 0
+
+table_manager:
+  chunk_tables_provisioning:
+    inactive_read_throughput: 0
+    inactive_write_throughput: 0
+    provisioned_read_throughput: 0
+    provisioned_write_throughput: 0
+  index_tables_provisioning:
+    inactive_read_throughput: 0
+    inactive_write_throughput: 0
+    provisioned_read_throughput: 0
+    provisioned_write_throughput: 0
+  retention_deletes_enabled: false
+  retention_period: 0
+```
+
+# Loki Deployment on Separate Mode
+To those who want to use Loki in production and persuit higher performance, deploying Loki on `Separate Mode` is a must.
+
+There're 4 things need to consider compared to Single Process Mode.
+1. The Ring Store:
+
+    You need to start a Consul server as the Ring store. It will store a consistent hash ring which consists of a list of tokens and ingesters. And Components like Distributors/Queriers rely on it to know which ingester is alive.
+    For more details, you can read doc [here](https://github.com/cortexproject/cortex/blob/master/docs/architecture.md#the-hash-ring).
+
+    Here's an example using docker to start a Consul server:
+    `docker run -e 'CHECKPOINT_DISABLE=1' -p 8300:8300 -p 8400:8400 -p 8500:8500 consul:0.7.1 agent -ui -server -client=0.0.0.0 -bootstrap`
+2. The Storage:
+
+    You need to think about using some index store and chunk store which can access from all above components.
+    For example, if you start above components in Kubernetes Pods, you may choose Amazon DynamoDB+S3, or choose boltdb+filesystem in a NFS storage like CephFS.
+3. The http_listen_port and grpc_listen_port:
+
+    If you start above components separately in different Kubernetes Pod or in different VM/BM, you don't need to worry about it.
+    If you start above components separately in your dev machine or one VM/BM, you need assign different port to them. Otherwise you will see errors like "address already in use".
+4. The startup command args
+
+    You need to specify the target for different component. 
+    For example, to start up Ingester only, using command like `./loki -config.file=/etc/loki/local-config.yaml -target=ingester`
+    Other components are similar.
+
+Below is the sample Loki Config for Separate Mode, please pay attention to the `ingester.lifecycler.ring.kvstore` part.
+```yaml
+uth_enabled: false
+
+server:
+  http_listen_port: 3100
+
+ingester:
+  lifecycler:
+    ring:
+      kvstore:
+        store: consul
+        consul:
+          host: consulhost:8500
+          prefix: ''
+          httpclienttimeout: 20s
+          consistentreads: true
+      replication_factor: 1
+  chunk_idle_period: 15m
+
+schema_config:
+  configs:
+  - from: 2018-04-15
+    store: boltdb
+    object_store: filesystem
+    schema: v9
+    index:
+      prefix: index_
+      period: 168h
+
+storage_config:
+  boltdb:
+    directory: /import/ceph/loki/index
+
+  filesystem:
+    directory: /import/ceph/loki/chunks
+
+limits_config:
+  enforce_metric_name: false
+  reject_old_samples: true
+  reject_old_samples_max_age: 168h
+
+chunk_store_config:
+  max_look_back_period: 0
+
+table_manager:
+  chunk_tables_provisioning:
+    inactive_read_throughput: 0
+    inactive_write_throughput: 0
+    provisioned_read_throughput: 0
+    provisioned_write_throughput: 0
+  index_tables_provisioning:
+    inactive_read_throughput: 0
+    inactive_write_throughput: 0
+    provisioned_read_throughput: 0
+    provisioned_write_throughput: 0
+  retention_deletes_enabled: false
+  retention_period: 0
+```