refactoring for later security changes

This commit is contained in:
Chris Lu
2015-01-05 14:20:04 -08:00
parent 878d1aa35e
commit a3e4145e8a
9 changed files with 230 additions and 57 deletions

36
note/security.txt Normal file
View File

@@ -0,0 +1,36 @@
Design for Seaweed-FS security
Design Objectives
Security can mean many different things. The original vision is that: if you have one machine lying around
somewhere with some disk space, it should be able to join your file system to contribute some disk space and
network bandwidth.
To achieve this purpose, the security should be able to:
1. Secure the inter-server communication. Only real cluster servers can join and communicate.
2. allow clients to securely write to volume servers
Non Objective
Multi-tenant support. Avoid filers or clients cross-updating files.
User specific access control.
Design Architect
master, and volume servers all talk securely via 2-way SSL for admin.
upon joining, master gives its secret key to volume servers.
filer or clients talk to master to get secret key, and use the key to generate JWT to write on volume server.
A side benefit:
a time limited read feature?
4. volume server needs to expose https ports
HTTP Connections
clear http
filer~>master, need to get a JWT from master
filer~>volume
2-way https
master~ssl~>volume
volume~ssl~>master
file uploading:
when volume server starts, it asks master for the secret key to decode JWT
when filer/clients wants to upload, master generate a JWT
filer~>volume(public port)
master~>volume(public port)