MyFavMirrors/minio
1
0
Fork 0
mirror of https://github.com/minio/minio.git synced 2024-12-26 07:05:55 -05:00
Code Issues Packages Projects Releases Wiki Activity
6df1e4a529
minio/docs/federation/lookup/README.md
Nitish 6df1e4a529 Add functionality to add old buckets to etcd on startup 
Buckets already present on a Minio server before it joins a
bucket federated deployment will now be added to etcd during
startup. In case of a bucket name collision, admin is informed
via Minio server console message.

Added configuration migration for configuration stored in etcd
backend.

Also, environment variables are updated and ListBucket path style
request is no longer forwarded.
2018-06-08 10:22:01 -07:00

1.8 KiB
Raw Blame History

Federation

There are primarily two types of federation

  • Bucket lookup from DNS
  • Bucket is shared across many clusters

This document will explain in detail about how to configure Minio supporting Bucket lookup

Federation (Bucket Lookup)

Bucket lookup federation requires two dependencies

  • etcd (for config, bucket SRV records)
  • coredns (for DNS management based on populated bucket SRV records)

Architecture

bucket-lookup

Run Multiple Clusters

cluster1

export MINIO_ETCD_ENDPOINTS="http://remote-etcd1:2379,http://remote-etcd2:4001"
export MINIO_DOMAIN=domain.com
export MINIO_PUBLIC_IP=44.35.2.1
minio server http://rack{1...4}.host{1...4}.domain.com/mnt/export{1...32}

cluster2

export MINIO_ETCD_ENDPOINTS="http://remote-etcd1:2379,http://remote-etcd2:4001"
export MINIO_DOMAIN=domain.com
export MINIO_PUBLIC_IP=44.35.2.2
minio server http://rack{5...8}.host{5...8}.domain.com/mnt/export{1...32}

In this configuration you can see MINIO_ETCD_ENDPOINTS points to the etcd backend which manages Minio's config.json and bucket SRV records. MINIO_DOMAIN indicates the domain suffix for the bucket which will be used to resolve bucket from DNS. For example if you have a bucket such as mybucket, the client can use now mybucket.domain.com to directly resolve to the right cluster. MINIO_PUBLIC_IP points to the public IP address where each cluster might be accessible, this is unique per each cluster.

NOTE: mybucket only exists on one cluster either cluster1 or cluster2 this is truly random and is decided by how domain.com gets resolved, if there is a round-robin DNS on domain.com then it is truly random which cluster might provision the bucket. This control is not provided to the client yet, but can be done based on the region parameter as supported by AWS S3 specification.