Ceph-Brag¶

Summary¶

Ceph-brag is going to be an anonymized cluster reporting tool designed to collect a "registry" of Ceph clusters for community knowledge. This data will be displayed on a public web page using UUID by default, but users can claim their cluster and publish information about ownership if they so desire.

Owners¶

• Loic Dachary <loic@dachary.org>
• Patrick McGarry (Inktank)
• Sebastien Han (eNovance)
• Sage Weil (Inktank)

Interested Parties¶

• Mike Dawson <mike.dawson@cloudapt.com>
• Haomai Wang <haomaiwang@gmail.com>
• Danny Al-Gaaf <danny.al-gaaf@bisect.de>

Current Status¶

• Use the http://wiki.ceph.com/Brag namespace for publication
• A teuthology task should be able to publish results if given proper credentials
• Setup consists of (declarative - e.g JSON/XML exportable) Topology, Hardware, OSs, Ceph (setup method, command/script steps and/or chef/puppet artifacts, final running ceph configs on node classes).
• Apdex (Application Performance Index) is an open standard developed by an alliance of companies. It defines a standard method for reporting and comparing the performance of software applications in computing. Its purpose is to convert measurements into insights about user satisfaction, by specifying a uniform way to analyze and report on the degree to which measured performance meets user expectations. http://en.wikipedia.org/wiki/Apdex
• What relationship with https://blueprints.launchpad.net/oslo/+spec/opt-in-stats-tracking ?
• Check the new osd metadata as a way to extract information https://github.com/ceph/ceph/pull/843
• lshw -xml to extract the hardware configuration
• CAS of the RAM https://github.com/enovance/edeploy/blob/master/build/sources/timings.c
• megacli on dell & hpacucli on HP ... to figure out more about disks

Detailed Description¶

Client side is a 'ceph brag' or 'ceph-brag' command. Generates a lump of JSON that is anonymous and sends it to brag.ceph.com (or similar). Includes:

• a unique identifier for the cluster. this is not the cluster fsid, but a new uuid, generated once and stored via the config-key interface, so that subsequent ceph-brag commands wil re-use the same id.
• cluster creation date
• number of osds, mons, mdss, pgs
• number of bytes, objects, pools
• number of bytes, ios read/written
• number of unique ips (hosts)?
• count of crush items by type (root, rack, host, osd)
• per-pool metadata
  • replica count
  • type (just rep for now, soon ec)
• os, kernel info (once available)
• hardware information (CPUs,RAM,Network ... maybe reduced to some basic anonymised data)
• ceph version(s)

On each brag, dump the current ownership information for the cluster. By default this is empty/undefined. ceph-brag options can be used to update the following fields:

• Name
• Organization
• Contact email
• Use-case (rgw, openstack, genomics, hpc, log archival, backups, whatever)

We need to make sure to expose no critical information to the public that could be a source to run e.g. exploits or DDoS attacks against a cluster. This is critical, otherwise no big company will ever expose any information via this tool. It could be even a problem to provide information abou the used ceph version.

Usage:

ceph-brag # generate brag json, dump to stdout
ceph-brag publish # post it!
ceph-brag update-metadata --name ... --organization ... --email ... --description ...
ceph-brag clear-metadata
ceph-brag unpublish --yes-i-am-shy

Server side is some WSGI or similar modern/cute REST endpoint. Simply logs the result to a database.

Work items¶

Coding tasks¶

1. ceph-brag
   1. generate all the json
2. ceph-brag publish
3. ceph-brag update-metadata
4. ceph-brag clear-metadata
5. ceph-brag unpublish
6. ceph-brag server
7. basic tool to summarize results
   1. number of clusters, bytes, objects
   2. os, ceph version histograms

Build / release tasks¶

1. deploy brag server to ceph.com

Documentation tasks¶

1. document security implications of ceph-brag
2. document how to obtain the public database
   1. programmatically, or by request from a human?