)]}'
{"specs/liberty/distributed-database-backend.rst":[{"author":{"_account_id":7725,"name":"David Stanek","email":"dstanek@dstanek.com","username":"dstanek"},"change_message_id":"009b03d12ecdca3316d568dd7a7efda5993b24ba","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"`bp distributed-db-backend \u003chttps://blueprints.launchpad.net/keystone/+spec/distributed-database-backend\u003e`_"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"The addition of an optional driver for keystone backends is proposed. This"},{"line_number":14,"context_line":"driver is intended to be based on a NoSQL database, specifically Cassandra."},{"line_number":15,"context_line":""},{"line_number":16,"context_line":"Problem Description"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_b9761ed5","line":13,"updated":"2015-06-17 21:41:27.000000000","message":"Keystone","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":31,"context_line":"To help with large deployments, it is proposed to add an optional NoSQL"},{"line_number":32,"context_line":"backend based on Cassandra to address both concerns of single-point-of-"},{"line_number":33,"context_line":"failure, and lack of ability to scale-out. The trade-off to be made for such a"},{"line_number":34,"context_line":"solution is to accept eventual consistency instead of strong consistency"},{"line_number":35,"context_line":"offered by existing solutions. This is in keeping with `OpenStack Desgin"},{"line_number":36,"context_line":"Tenets \u003chttps://wiki.openstack.org/wiki/BasicDesignTenets\u003e`."},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_a69624ee","line":34,"updated":"2015-05-04 14:37:13.000000000","message":"so it\u0027s equivalent to master-slave MySQL when it comes to point of failure and scale out considerations?","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"8a38804ead80bfef76cf03bcdf4449c80c007bd0","unresolved":false,"context_lines":[{"line_number":31,"context_line":"To help with large deployments, it is proposed to add an optional NoSQL"},{"line_number":32,"context_line":"backend based on Cassandra to address both concerns of single-point-of-"},{"line_number":33,"context_line":"failure, and lack of ability to scale-out. The trade-off to be made for such a"},{"line_number":34,"context_line":"solution is to accept eventual consistency instead of strong consistency"},{"line_number":35,"context_line":"offered by existing solutions. This is in keeping with `OpenStack Desgin"},{"line_number":36,"context_line":"Tenets \u003chttps://wiki.openstack.org/wiki/BasicDesignTenets\u003e`."},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_3c48e086","line":34,"in_reply_to":"3acd31a7_41105aa8","updated":"2015-05-04 16:39:24.000000000","message":"s/don\u0027t/doesn\u0027t","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"e64369d99c78e9303e7861c650287bb9c90465bb","unresolved":false,"context_lines":[{"line_number":31,"context_line":"To help with large deployments, it is proposed to add an optional NoSQL"},{"line_number":32,"context_line":"backend based on Cassandra to address both concerns of single-point-of-"},{"line_number":33,"context_line":"failure, and lack of ability to scale-out. The trade-off to be made for such a"},{"line_number":34,"context_line":"solution is to accept eventual consistency instead of strong consistency"},{"line_number":35,"context_line":"offered by existing solutions. This is in keeping with `OpenStack Desgin"},{"line_number":36,"context_line":"Tenets \u003chttps://wiki.openstack.org/wiki/BasicDesignTenets\u003e`."},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_41105aa8","line":34,"in_reply_to":"3acd31a7_a69624ee","updated":"2015-05-04 16:38:27.000000000","message":"No it\u0027s not. The above line says that the concerns associated with MySQL are, it\u0027s a SPOF and it don\u0027t scale out. The Cassandra solution addresses both of these problems.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":32,"context_line":"backend based on Cassandra to address both concerns of single-point-of-"},{"line_number":33,"context_line":"failure, and lack of ability to scale-out. The trade-off to be made for such a"},{"line_number":34,"context_line":"solution is to accept eventual consistency instead of strong consistency"},{"line_number":35,"context_line":"offered by existing solutions. This is in keeping with `OpenStack Desgin"},{"line_number":36,"context_line":"Tenets \u003chttps://wiki.openstack.org/wiki/BasicDesignTenets\u003e`."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Next, the problem of disaster recovery needs to be addressed by any approach"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_461070a8","line":35,"updated":"2015-05-04 14:37:13.000000000","message":"Desgin -\u003e Design","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":38,"context_line":"Next, the problem of disaster recovery needs to be addressed by any approach"},{"line_number":39,"context_line":"taken. MySQL based deployments typically have off-site replication for this"},{"line_number":40,"context_line":"purpose with manual steps to failover to a different site, should disaster"},{"line_number":41,"context_line":"strike the primary one. The Cassandra based NoSQL solution naturally extends"},{"line_number":42,"context_line":"to the disaster recovery scenario as well."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Lastly, there is an admittedly minor benefit of this approach for moderate"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_06515862","line":41,"updated":"2015-05-04 14:37:13.000000000","message":"how does the NoSQL solution naturally extend to the disaster recovery scenario? Please elaborate.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":7725,"name":"David Stanek","email":"dstanek@dstanek.com","username":"dstanek"},"change_message_id":"009b03d12ecdca3316d568dd7a7efda5993b24ba","unresolved":false,"context_lines":[{"line_number":38,"context_line":"Next, the problem of disaster recovery needs to be addressed by any approach"},{"line_number":39,"context_line":"taken. MySQL based deployments typically have off-site replication for this"},{"line_number":40,"context_line":"purpose with manual steps to failover to a different site, should disaster"},{"line_number":41,"context_line":"strike the primary one. The Cassandra based NoSQL solution naturally extends"},{"line_number":42,"context_line":"to the disaster recovery scenario as well."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Lastly, there is an admittedly minor benefit of this approach for moderate"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_bc392cd9","line":41,"in_reply_to":"1af86dd1_695ef8ea","updated":"2015-06-17 21:41:27.000000000","message":"I agree. OTOH this is just how Cassandra works.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"e64369d99c78e9303e7861c650287bb9c90465bb","unresolved":false,"context_lines":[{"line_number":38,"context_line":"Next, the problem of disaster recovery needs to be addressed by any approach"},{"line_number":39,"context_line":"taken. MySQL based deployments typically have off-site replication for this"},{"line_number":40,"context_line":"purpose with manual steps to failover to a different site, should disaster"},{"line_number":41,"context_line":"strike the primary one. The Cassandra based NoSQL solution naturally extends"},{"line_number":42,"context_line":"to the disaster recovery scenario as well."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Lastly, there is an admittedly minor benefit of this approach for moderate"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_616eb602","line":41,"in_reply_to":"3acd31a7_06515862","updated":"2015-05-04 16:38:27.000000000","message":"You could have a Cassandra cluster running in one of your DR site and have your in-use Cassandra cluster replicate data asynchronously to that site. You don\u0027t need to use an extra replication solution to do this. Cassandra does this on its own.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":5046,"name":"Lance Bragstad","email":"lbragstad@redhat.com","username":"ldbragst"},"change_message_id":"3b7a2b1c64aa3b05fb84a300775895fbafce57ca","unresolved":false,"context_lines":[{"line_number":38,"context_line":"Next, the problem of disaster recovery needs to be addressed by any approach"},{"line_number":39,"context_line":"taken. MySQL based deployments typically have off-site replication for this"},{"line_number":40,"context_line":"purpose with manual steps to failover to a different site, should disaster"},{"line_number":41,"context_line":"strike the primary one. The Cassandra based NoSQL solution naturally extends"},{"line_number":42,"context_line":"to the disaster recovery scenario as well."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Lastly, there is an admittedly minor benefit of this approach for moderate"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1af86dd1_695ef8ea","line":41,"in_reply_to":"3acd31a7_616eb602","updated":"2015-06-12 19:41:47.000000000","message":"Explaining that in the spec would be good.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":47,"context_line":"milliseconds of latency at a minimum. In such a scenario, user identity is"},{"line_number":48,"context_line":"expected to carry over across regions. The proposed solution handles this"},{"line_number":49,"context_line":"scenario without the need for federation since replication across regions"},{"line_number":50,"context_line":"- without necessarily increasing latencies for each request - is a natural"},{"line_number":51,"context_line":"part of the proposed solution. Of course this approach is probably"},{"line_number":52,"context_line":"not practical beyond two or three regions."},{"line_number":53,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_86c0c8e1","line":50,"updated":"2015-05-04 14:37:13.000000000","message":"how so? Please elaborate.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"e64369d99c78e9303e7861c650287bb9c90465bb","unresolved":false,"context_lines":[{"line_number":47,"context_line":"milliseconds of latency at a minimum. In such a scenario, user identity is"},{"line_number":48,"context_line":"expected to carry over across regions. The proposed solution handles this"},{"line_number":49,"context_line":"scenario without the need for federation since replication across regions"},{"line_number":50,"context_line":"- without necessarily increasing latencies for each request - is a natural"},{"line_number":51,"context_line":"part of the proposed solution. Of course this approach is probably"},{"line_number":52,"context_line":"not practical beyond two or three regions."},{"line_number":53,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_bc4ab0aa","line":50,"in_reply_to":"3acd31a7_86c0c8e1","updated":"2015-05-04 16:38:27.000000000","message":"The database query in each region go to only the local Cassandra cluster and Cassandra replicates the data across regions asynchronously.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":7725,"name":"David Stanek","email":"dstanek@dstanek.com","username":"dstanek"},"change_message_id":"009b03d12ecdca3316d568dd7a7efda5993b24ba","unresolved":false,"context_lines":[{"line_number":47,"context_line":"milliseconds of latency at a minimum. In such a scenario, user identity is"},{"line_number":48,"context_line":"expected to carry over across regions. The proposed solution handles this"},{"line_number":49,"context_line":"scenario without the need for federation since replication across regions"},{"line_number":50,"context_line":"- without necessarily increasing latencies for each request - is a natural"},{"line_number":51,"context_line":"part of the proposed solution. Of course this approach is probably"},{"line_number":52,"context_line":"not practical beyond two or three regions."},{"line_number":53,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_7cb40434","line":50,"in_reply_to":"3acd31a7_bc4ab0aa","updated":"2015-06-17 21:41:27.000000000","message":"Again...add this to the doc. The idea is to have one big database distributes across regions and have a region specific Keystone talk to the DB instance in its region.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":5046,"name":"Lance Bragstad","email":"lbragstad@redhat.com","username":"ldbragst"},"change_message_id":"3b7a2b1c64aa3b05fb84a300775895fbafce57ca","unresolved":false,"context_lines":[{"line_number":49,"context_line":"scenario without the need for federation since replication across regions"},{"line_number":50,"context_line":"- without necessarily increasing latencies for each request - is a natural"},{"line_number":51,"context_line":"part of the proposed solution. Of course this approach is probably"},{"line_number":52,"context_line":"not practical beyond two or three regions."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"Proposed Change"},{"line_number":55,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1af86dd1_e920a84b","line":52,"updated":"2015-06-12 19:41:47.000000000","message":"With a distributed backend that handles replication, Galera for example, it is recommended that you deploy in an architecture that can handle quorum. How does Cassandra handle this and will it require Keystone to be deployed to a specific reference architecture for it to work?","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":80,"context_line":"None."},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"Performance Impact"},{"line_number":83,"context_line":"------------------"},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"None."},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_a64bc486","line":83,"updated":"2015-05-04 14:37:13.000000000","message":"Compared to which current case? And is there some evidence you can point to here, showing that the performance will be equivalent?","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"e64369d99c78e9303e7861c650287bb9c90465bb","unresolved":false,"context_lines":[{"line_number":80,"context_line":"None."},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"Performance Impact"},{"line_number":83,"context_line":"------------------"},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"None."},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_bce1909c","line":83,"in_reply_to":"3acd31a7_a64bc486","updated":"2015-05-04 16:38:27.000000000","message":"Compared to MySQL. We are working on a performance test with Cassandra implementation. We don\u0027t have numbers yet.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":7725,"name":"David Stanek","email":"dstanek@dstanek.com","username":"dstanek"},"change_message_id":"cb2be569d2b5bc1b08a5ab0ce3f75e6b889673fe","unresolved":false,"context_lines":[{"line_number":88,"context_line":"---------------------"},{"line_number":89,"context_line":""},{"line_number":90,"context_line":"This is going to be another driver. Deployers are free not to use it and"},{"line_number":91,"context_line":"use default MySQL driver. If this driver is used then a Cassandra cluster has"},{"line_number":92,"context_line":"to be deployed."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Add a new paramter cassandra_node_ips in database section."}],"source_content_type":"text/x-rst","patch_set":2,"id":"da20952f_880c5aee","line":91,"updated":"2015-08-28 20:39:39.000000000","message":"The default is SQL not MySQL.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":10608,"name":"Matthew Edmonds","email":"edmondsw@us.ibm.com","username":"edmondsw"},"change_message_id":"9804bfa79739fa3d9954c0a75e5c1a75a5d0530e","unresolved":false,"context_lines":[{"line_number":91,"context_line":"use default MySQL driver. If this driver is used then a Cassandra cluster has"},{"line_number":92,"context_line":"to be deployed."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Add a new paramter cassandra_node_ips in database section."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"This driver is going to reside in backend directory of each keystone backend."},{"line_number":97,"context_line":"So the driver should be changed for each backend if the new driver is used."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_c6548022","line":94,"updated":"2015-05-04 14:37:13.000000000","message":"paramter -\u003e parameter\n\ncassandra_node_ips -\u003e that makes this cassandra-specific... should it be more generic, e.g. nosql_node_ips?","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":11428,"name":"Ajaya Agrawal","email":"ajku.agr@gmail.com","username":"ajayaa"},"change_message_id":"e64369d99c78e9303e7861c650287bb9c90465bb","unresolved":false,"context_lines":[{"line_number":91,"context_line":"use default MySQL driver. If this driver is used then a Cassandra cluster has"},{"line_number":92,"context_line":"to be deployed."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Add a new paramter cassandra_node_ips in database section."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"This driver is going to reside in backend directory of each keystone backend."},{"line_number":97,"context_line":"So the driver should be changed for each backend if the new driver is used."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_5ca03c43","line":94,"in_reply_to":"3acd31a7_c6548022","updated":"2015-05-04 16:38:27.000000000","message":"We are writing a Cassandra driver. So Cassandra specific parameter.","commit_id":"45968833356fd8d0154011755a2170b88caeb191"},{"author":{"_account_id":5046,"name":"Lance Bragstad","email":"lbragstad@redhat.com","username":"ldbragst"},"change_message_id":"3b7a2b1c64aa3b05fb84a300775895fbafce57ca","unresolved":false,"context_lines":[{"line_number":125,"context_line":"* Resource"},{"line_number":126,"context_line":"* Credential"},{"line_number":127,"context_line":"* Trust"},{"line_number":128,"context_line":"* Assignment"},{"line_number":129,"context_line":""},{"line_number":130,"context_line":"Dependencies"},{"line_number":131,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1af86dd1_e9e78810","line":128,"updated":"2015-06-12 19:41:47.000000000","message":"What about the catalog backend?","commit_id":"45968833356fd8d0154011755a2170b88caeb191"}]}
