Ceph

h1. 2G - RADOS namespaces CRUSH language extension CRUSH library

h3. Live Pad

The live pads can be found here: "[pad 1]":http://pad.ceph.com/p/rados_namespaces "[pad 2]":http://pad.ceph.com/p/extend_crush_rule_language "[pad 3]":http://pad.ceph.com/p/create_crush_library

h3. Summit Snapshot

h4. RADOS Namespaces

Coding tasks

# protocol update for MOSDOp, MOSDOpReply to add namespace fields

# librados API changes: new versions of a handful of calls to add the new fields

## [aio_]operate (add C interface)

# os/FileStore changes

## almost nothing; hobject_t has a ns field.

## extend object file naming scheme and omap key naming scheme as needed (this may already be handled with the last hobject_t rev we did last year) [or maybe they already have them]

# extend OSDCap grammar and associated tests

## allow pool <foo> namespace <bar>

# OSD changes: add ns argument every that object is used

## anywhere there is an sobject_t... hopefully none or very few

# manage compatibility: make OSD error out gracefully if not all replicas support namespaces, prevent downgrade, etc.  This may require a global osdmap flag indicating that namespaces are supported.

## once all osds in the OSDMap have the feature, we set the flag in the OSDMap

## MOSDBoot needs to communicate features

# extend librados list_objects() to have a ns field (along with object name and key)

# Update RBD to allow images to live in namespaces.

## inventing a convenience syntax? pool/namespace/image or pool/image ?

## disallow / and @ and : in image names

# Add namespace argument to rados cli tool

# make rados bench write into a namespace!

# add rados tool command to remove a namespace

Testing tasks

# extend librados API tests

# extend existing filestore and osd stress tests to stress the namespace code

## make radosmodel test list_objects() [at a snapshot!!]

# extend osd cap tests

Documentation tasks

# Document OSDCap grammar and semantics

# Document namespace purpose, uses, and limitations

# Document how namespaces can be leveraged by OpenStack RBD users

Things we do support:

# unbounded # of namespaces

# security limits by namespace

Things we don't support:

# efficient listing of namespace

# efficient listing of namespace contexts

# per-namespace stats

h4. CRUSH Language Extension

Coding tasks

# extend rule parser

# extend mapping function to implement new rule language

# port to kernel implementation

# define apppriate compatibility bits and integrate into existing Ceph compat checks

Documentation tasks

# document new rule language

Possibilities:

# a stack to supplement the working vector

# chooseleaf would take a min # of unique items as well as total count of item

## needs some way to intelligently separate items beneath that point when there are dups

# sum over map of (type -> weight) instead of just weight

# block bucket summation

Expand crush tester/simulator

Comments:

* You might want to look into the route-map user interface of Cisco routers. I thought it handles significantly complex policies the same level as we want in this case.

h4. CRUSH Library

Coding tasks

# clean up existing API for consistency, argument types, etc.

# define new API (C and/or C++)

# refactor CRUSH code

# adjust Ceph code to utilize new API

# build (at least a basic) unit test suite for the new API

Build / release tasks

# adjust ceph build

# package libcrush separately (deb, rpm)

Documentation tasks

# document new crush API