I have gone over the first 48 of the tickets listed in the Google Code at http://code.google.com/p/project-voldemort/issues/list Most of them are fixed already or relate to non-functional aspects like code cleanup or refer to code paths which are not the default (i.e ThreadPool based routed store)
I have moved over things that are either legit bug fixes or features that fit into our roadmap. Please feel free to reopen them on github if some issue I skipped over is very important for your use cases.
Below is the list of tickets and rationale.
2 Provide publish/subscribe mechanism move; Should be addressed by Event log; 9 Make property names consistent move; 24 Refactoring of JDBC-based storage engines skip; but will checkin any changes we make to MySql engine based on storage investigation 44 Make separate client jar skip; Non critical 50 Json Serialization should support "not-null" feature. skip; We are improving AVRO support. Moving towards standards 61 Finish QueryServlet skip, Not in our roadmap 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded tests in Hudson DONE, Hudson runs a bunch of integration like tests. 87 Partition-Rebalance sub task 5: Partition Rebalancing User Interface skip; Not in our roadmap 91 Implement RoutedStoreTest.testObsoleteMasterFails move 105 JsonReader.java reads int not long move 117 Timeout configuration for getAll DONE, in 0.96 119 Implement daemon functionality using Akuma skip 120 WAR dependencies are a bit of a mess skip 121 Run C++ tests in Hudson skip 128 Allow required reads to be specified during get skip, We are also thinking about changes to client api including async apis, thin client. So this will be considered in that light 130 Use variants to improve space usage of serialized VectorClock DONE 132 Cacti-formatted Output from Status Servlet skip, Not in our roadmap 140 Port RoutedStore.put() fixes to RoutedStore.delete() DONE 141 AbstractSocketPoolTest.startTest calls JUnit assert methods outside main thread move 160 Move vector-clock resolution lower in the processing stack DONE 162 Put should throw unique exceptions for its error cases (no master, too few replicas, too few successes) skip, I think the current message is descriptive 166 The enabled storage engines should be derived automatically from the stores.xml file move 169 Berkeley DB File Utilization stats should be exposed DONE 174 VoldemortConfig validation should be made uniform move 179 Study the impact of opening bdb environment per store partition Skip, With Partition based scans currently being implemented, we may not need this 180 Voldemort Consistency mechanisms Will not be executing specifically on this item. But these options will be evaluated when we get to exploring server side routing and event log feature 182 BdbStorageEngine#put may not commit or abort transaction DONE 183 AdminClient VoldemortFilter over the wire class transfer issue. skip, low priority considering filters are not very commonly used 186 truncate() failing for bdbStorageEngine. DONE 189 VersionIncrementingStore is not updating the Versioned object after a successful put skip, test only change 193 Scheduled jobs fail with no exception skip, logging issue is fixed with a big try-catch in current streaming jobs 197 Notification mechanism for changes in the cluster node membership skip 213 Add StoreClient.contains skip, There is an open pull request from roshan. Duplicate. 214 Datacenter awareness DONE 220 Add StoreClient.deleteAll(keys) skip 221 Make statistics be sliding window based instead of being reset every X seconds DONE we have implemented our histograms which does this 224 Need to differentiate between client and server timeout errors (different exceptions) DONE 225 Improved CacheStorageConfiguration skip 227 Remove calls to deprecated API in BDB4 DONE 232 voldemort binds to all interfaces instead of binding to just the host defined in the cluster.xml file for that server move 235 Performance Issues/Timeouts with NIO and Lots of Clients Skip; We use NIO and seems to be doing fine now. Also, I experimented with separation of network and disk IO in the server. What I found was that the performance actually suffered due to thread contention when disk was 'cheap' 237 Voldemort should have way of limiting the size of keys and values in a store move 238 Protocol Versioning Negotiation Needs to be Bi-directional skip, not critical 239 Voldemort should not support localhost in cluster.xml files skip 241 SocketStore redesign DONE, Jay Wylie has been working on this. If all goes well, expect some improvements in next releases 242 RoutedStore should abandon earlier if not enough nodes remain on PUT DONE all of this is already fixed/differently done in the pipeline routed store 243 Files generated by ConfigurationStorageEngine should not have xml extension unless they're actually XML skip
> I have gone over the first 48 of the tickets listed in the Google Code at http://code.google.com/p/project-voldemort/issues/list > Most of them are fixed already or relate to non-functional aspects like
> code cleanup or refer to code paths which are not the default (i.e
> ThreadPool based routed store)
> I have moved over things that are either legit bug fixes or features that
> fit into our roadmap. Please feel free to reopen them on github if some
> issue I skipped over is very important for your use cases.
> Below is the list of tickets and rationale.
> 2 Provide publish/subscribe mechanism move; Should be addressed by
> Event log;
> 9 Make property names consistent move; 24 Refactoring of JDBC-based
> storage engines skip; but will checkin any changes we make to MySql
> engine based on storage investigation
> 44 Make separate client jar skip; Non critical
> 50 Json Serialization should support "not-null" feature. skip; We are
> improving AVRO support. Moving towards standards 61 Finish QueryServlet skip,
> Not in our roadmap
> 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded tests
> in Hudson DONE, Hudson runs a bunch of integration like tests.
> 87 Partition-Rebalance sub task 5: Partition Rebalancing User Interface skip;
> Not in our roadmap 91 Implement RoutedStoreTest.testObsoleteMasterFails
> move 105 JsonReader.java reads int not long move 117 Timeout
> configuration for getAll DONE, in 0.96 119 Implement daemon
> functionality using Akuma skip 120 WAR dependencies are a bit of a mess
> skip 121 Run C++ tests in Hudson skip 128 Allow required reads to be
> specified during get skip, We are also thinking about changes to client
> api including async apis, thin client. So this will be considered in that
> light 130 Use variants to improve space usage of serialized VectorClock
> DONE 132 Cacti-formatted Output from Status Servlet skip, Not in our
> roadmap 140 Port RoutedStore.put() fixes to RoutedStore.delete() DONE
> 141 AbstractSocketPoolTest.startTest calls JUnit assert methods outside
> main thread move 160 Move vector-clock resolution lower in the
> processing stack DONE 162 Put should throw unique exceptions for its
> error cases (no master, too few replicas, too few successes) skip, I
> think the current message is descriptive 166 The enabled storage
> engines should be derived automatically from the stores.xml file move 169 Berkeley
> DB File Utilization stats should be exposed DONE 174 VoldemortConfig
> validation should be made uniform move 179 Study the impact of opening
> bdb environment per store partition Skip, With Partition based scans
> currently being implemented, we may not need this 180 Voldemort
> Consistency mechanisms Will not be executing specifically on this item.
> But these options will be evaluated when we get to exploring server side
> routing and event log feature 182 BdbStorageEngine#put may not commit or
> abort transaction DONE 183 AdminClient VoldemortFilter over the wire
> class transfer issue. skip, low priority considering filters are not very
> commonly used 186 truncate() failing for bdbStorageEngine. DONE 189 VersionIncrementingStore
> is not updating the Versioned object after a successful put skip, test
> only change 193 Scheduled jobs fail with no exception skip, logging
> issue is fixed with a big try-catch in current streaming jobs 197 Notification
> mechanism for changes in the cluster node membership skip 213 Add
> StoreClient.contains skip, There is an open pull request from roshan.
> Duplicate. 214 Datacenter awareness DONE 220 Add
> StoreClient.deleteAll(keys) skip 221 Make statistics be sliding window
> based instead of being reset every X seconds DONE we have implemented our
> histograms which does this 224 Need to differentiate between client and
> server timeout errors (different exceptions) DONE 225 Improved
> CacheStorageConfiguration skip 227 Remove calls to deprecated API in BDB4
> DONE 232 voldemort binds to all interfaces instead of binding to just
> the host defined in the cluster.xml file for that server move 235 Performance
> Issues/Timeouts with NIO and Lots of Clients Skip; We use NIO and seems
> to be doing fine now. Also, I experimented with separation of network and
> disk IO in the server. What I found was that the performance actually
> suffered due to thread contention when disk was 'cheap' 237 Voldemort
> should have way of limiting the size of keys and values in a store move
> 238 Protocol Versioning Negotiation Needs to be Bi-directional skip, not
> critical 239 Voldemort should not support localhost in cluster.xml files
> skip 241 SocketStore redesign DONE, Jay Wylie has been working on this.
> If all goes well, expect some improvements in next releases 242 RoutedStore
> should abandon earlier if not enough nodes remain on PUT DONE all of this
> is already fixed/differently done in the pipeline routed store 243 Files
> generated by ConfigurationStorageEngine should not have xml extension
> unless they're actually XML skip
On Wednesday, 19 September 2012 19:06:41 UTC-4, ctasada wrote:
> Great work!
> On Wed, Sep 19, 2012 at 12:30 PM, Vinoth Chandar <mail.vino...@gmail.com<javascript:> > > wrote:
>> Guys,
>> I have gone over the first 48 of the tickets listed in the Google Code at http://code.google.com/p/project-voldemort/issues/list >> Most of them are fixed already or relate to non-functional aspects like >> code cleanup or refer to code paths which are not the default (i.e >> ThreadPool based routed store)
>> I have moved over things that are either legit bug fixes or features that >> fit into our roadmap. Please feel free to reopen them on github if some >> issue I skipped over is very important for your use cases.
>> Below is the list of tickets and rationale.
>> 2 Provide publish/subscribe mechanism move; Should be addressed by >> Event log; >> 9 Make property names consistent move; 24 Refactoring of JDBC-based >> storage engines skip; but will checkin any changes we make to MySql >> engine based on storage investigation >> 44 Make separate client jar skip; Non critical >> 50 Json Serialization should support "not-null" feature. skip; We are >> improving AVRO support. Moving towards standards 61 Finish QueryServlet skip, >> Not in our roadmap >> 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded tests >> in Hudson DONE, Hudson runs a bunch of integration like tests. >> 87 Partition-Rebalance sub task 5: Partition Rebalancing User Interface skip; >> Not in our roadmap 91 Implement RoutedStoreTest.testObsoleteMasterFails >> move 105 JsonReader.java reads int not long move 117 Timeout >> configuration for getAll DONE, in 0.96 119 Implement daemon >> functionality using Akuma skip 120 WAR dependencies are a bit of a mess >> skip 121 Run C++ tests in Hudson skip 128 Allow required reads to be >> specified during get skip, We are also thinking about changes to client >> api including async apis, thin client. So this will be considered in that >> light 130 Use variants to improve space usage of serialized VectorClock >> DONE 132 Cacti-formatted Output from Status Servlet skip, Not in our >> roadmap 140 Port RoutedStore.put() fixes to RoutedStore.delete() DONE >> 141 AbstractSocketPoolTest.startTest calls JUnit assert methods outside >> main thread move 160 Move vector-clock resolution lower in the >> processing stack DONE 162 Put should throw unique exceptions for its >> error cases (no master, too few replicas, too few successes) skip, I >> think the current message is descriptive 166 The enabled storage >> engines should be derived automatically from the stores.xml file move >> 169 Berkeley DB File Utilization stats should be exposed DONE 174 VoldemortConfig >> validation should be made uniform move 179 Study the impact of opening >> bdb environment per store partition Skip, With Partition based scans >> currently being implemented, we may not need this 180 Voldemort >> Consistency mechanisms Will not be executing specifically on this item. >> But these options will be evaluated when we get to exploring server side >> routing and event log feature 182 BdbStorageEngine#put may not commit >> or abort transaction DONE 183 AdminClient VoldemortFilter over the wire >> class transfer issue. skip, low priority considering filters are not >> very commonly used 186 truncate() failing for bdbStorageEngine. DONE >> 189 VersionIncrementingStore is not updating the Versioned object after >> a successful put skip, test only change 193 Scheduled jobs fail with no >> exception skip, logging issue is fixed with a big try-catch in current >> streaming jobs 197 Notification mechanism for changes in the cluster >> node membership skip 213 Add StoreClient.contains skip, There is an >> open pull request from roshan. Duplicate. 214 Datacenter awareness DONE >> 220 Add StoreClient.deleteAll(keys) skip 221 Make statistics be sliding >> window based instead of being reset every X seconds DONE we have >> implemented our histograms which does this 224 Need to differentiate >> between client and server timeout errors (different exceptions) DONE 225 Improved >> CacheStorageConfiguration skip 227 Remove calls to deprecated API in >> BDB4 DONE 232 voldemort binds to all interfaces instead of binding to >> just the host defined in the cluster.xml file for that server move 235 Performance >> Issues/Timeouts with NIO and Lots of Clients Skip; We use NIO and seems >> to be doing fine now. Also, I experimented with separation of network and >> disk IO in the server. What I found was that the performance actually >> suffered due to thread contention when disk was 'cheap' 237 Voldemort >> should have way of limiting the size of keys and values in a store move >> 238 Protocol Versioning Negotiation Needs to be Bi-directional skip, not >> critical 239 Voldemort should not support localhost in cluster.xml files >> skip 241 SocketStore redesign DONE, Jay Wylie has been working on this. >> If all goes well, expect some improvements in next releases 242 RoutedStore >> should abandon earlier if not enough nodes remain on PUT DONE all of >> this is already fixed/differently done in the pipeline routed store 243 Files >> generated by ConfigurationStorageEngine should not have xml extension >> unless they're actually XML skip
>> I will be doing the second batch in a few weeks
On Friday, September 21, 2012 3:42:50 PM UTC+2, Antony Stubbs wrote:
> Cool, but you should also change the message on the google code page that > says it's still used for issue tracking.
> On Wednesday, 19 September 2012 19:06:41 UTC-4, ctasada wrote:
>> Great work!
>> On Wed, Sep 19, 2012 at 12:30 PM, Vinoth Chandar <mail.vino...@gmail.com>wrote:
>>> Guys,
>>> I have gone over the first 48 of the tickets listed in the Google Code >>> at http://code.google.com/p/project-voldemort/issues/list Most of them >>> are fixed already or relate to non-functional aspects like code cleanup or >>> refer to code paths which are not the default (i.e ThreadPool based routed >>> store)
>>> I have moved over things that are either legit bug fixes or features >>> that fit into our roadmap. Please feel free to reopen them on github if >>> some issue I skipped over is very important for your use cases.
>>> Below is the list of tickets and rationale.
>>> 2 Provide publish/subscribe mechanism move; Should be addressed by >>> Event log; >>> 9 Make property names consistent move; 24 Refactoring of JDBC-based >>> storage engines skip; but will checkin any changes we make to MySql >>> engine based on storage investigation >>> 44 Make separate client jar skip; Non critical >>> 50 Json Serialization should support "not-null" feature. skip; We >>> are improving AVRO support. Moving towards standards 61 Finish >>> QueryServlet skip, Not in our roadmap >>> 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded >>> tests in Hudson DONE, Hudson runs a bunch of integration like tests. >>> 87 Partition-Rebalance sub task 5: Partition Rebalancing User Interface skip; >>> Not in our roadmap 91 Implement RoutedStoreTest.testObsoleteMasterFails >>> move 105 JsonReader.java reads int not long move 117 Timeout >>> configuration for getAll DONE, in 0.96 119 Implement daemon >>> functionality using Akuma skip 120 WAR dependencies are a bit of a mess >>> skip 121 Run C++ tests in Hudson skip 128 Allow required reads to be >>> specified during get skip, We are also thinking about changes to client >>> api including async apis, thin client. So this will be considered in that >>> light 130 Use variants to improve space usage of serialized VectorClock >>> DONE 132 Cacti-formatted Output from Status Servlet skip, Not in our >>> roadmap 140 Port RoutedStore.put() fixes to RoutedStore.delete() DONE >>> 141 AbstractSocketPoolTest.startTest calls JUnit assert methods outside >>> main thread move 160 Move vector-clock resolution lower in the >>> processing stack DONE 162 Put should throw unique exceptions for its >>> error cases (no master, too few replicas, too few successes) skip, I >>> think the current message is descriptive 166 The enabled storage >>> engines should be derived automatically from the stores.xml file move >>> 169 Berkeley DB File Utilization stats should be exposed DONE 174 VoldemortConfig >>> validation should be made uniform move 179 Study the impact of opening >>> bdb environment per store partition Skip, With Partition based scans >>> currently being implemented, we may not need this 180 Voldemort >>> Consistency mechanisms Will not be executing specifically on this >>> item. But these options will be evaluated when we get to exploring server >>> side routing and event log feature 182 BdbStorageEngine#put may not >>> commit or abort transaction DONE 183 AdminClient VoldemortFilter over >>> the wire class transfer issue. skip, low priority considering filters >>> are not very commonly used 186 truncate() failing for bdbStorageEngine. >>> DONE 189 VersionIncrementingStore is not updating the Versioned object >>> after a successful put skip, test only change 193 Scheduled jobs fail >>> with no exception skip, logging issue is fixed with a big try-catch in >>> current streaming jobs 197 Notification mechanism for changes in the >>> cluster node membership skip 213 Add StoreClient.contains skip, There >>> is an open pull request from roshan. Duplicate. 214 Datacenter >>> awareness DONE 220 Add StoreClient.deleteAll(keys) skip 221 Make >>> statistics be sliding window based instead of being reset every X seconds DONE >>> we have implemented our histograms which does this 224 Need to >>> differentiate between client and server timeout errors (different >>> exceptions) DONE 225 Improved CacheStorageConfiguration skip 227 Remove >>> calls to deprecated API in BDB4 DONE 232 voldemort binds to all >>> interfaces instead of binding to just the host defined in the cluster.xml >>> file for that server move 235 Performance Issues/Timeouts with NIO and >>> Lots of Clients Skip; We use NIO and seems to be doing fine now. Also, >>> I experimented with separation of network and disk IO in the server. What I >>> found was that the performance actually suffered due to thread contention >>> when disk was 'cheap' 237 Voldemort should have way of limiting the >>> size of keys and values in a store move 238 Protocol Versioning >>> Negotiation Needs to be Bi-directional skip, not critical 239 Voldemort >>> should not support localhost in cluster.xml files skip 241 SocketStore >>> redesign DONE, Jay Wylie has been working on this. If all goes well, >>> expect some improvements in next releases 242 RoutedStore should >>> abandon earlier if not enough nodes remain on PUT DONE all of this is >>> already fixed/differently done in the pipeline routed store 243 Files >>> generated by ConfigurationStorageEngine should not have xml extension >>> unless they're actually XML skip
>>> I will be doing the second batch in a few weeks
> On Friday, September 21, 2012 3:42:50 PM UTC+2, Antony Stubbs wrote:
>> Cool, but you should also change the message on the google code page that >> says it's still used for issue tracking.
>> On Wednesday, 19 September 2012 19:06:41 UTC-4, ctasada wrote:
>>> Great work!
>>> On Wed, Sep 19, 2012 at 12:30 PM, Vinoth Chandar <mail.vino...@gmail.com >>> > wrote:
>>>> Guys,
>>>> I have gone over the first 48 of the tickets listed in the Google Code >>>> at http://code.google.com/p/project-voldemort/issues/list Most of them >>>> are fixed already or relate to non-functional aspects like code cleanup or >>>> refer to code paths which are not the default (i.e ThreadPool based routed >>>> store)
>>>> I have moved over things that are either legit bug fixes or features >>>> that fit into our roadmap. Please feel free to reopen them on github if >>>> some issue I skipped over is very important for your use cases.
>>>> Below is the list of tickets and rationale.
>>>> 2 Provide publish/subscribe mechanism move; Should be addressed by >>>> Event log; >>>> 9 Make property names consistent move; 24 Refactoring of JDBC-based >>>> storage engines skip; but will checkin any changes we make to MySql >>>> engine based on storage investigation >>>> 44 Make separate client jar skip; Non critical >>>> 50 Json Serialization should support "not-null" feature. skip; We >>>> are improving AVRO support. Moving towards standards 61 Finish >>>> QueryServlet skip, Not in our roadmap >>>> 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded >>>> tests in Hudson DONE, Hudson runs a bunch of integration like tests. >>>> 87 Partition-Rebalance sub task 5: Partition Rebalancing User >>>> Interface skip; Not in our roadmap 91 Implement >>>> RoutedStoreTest.testObsoleteMasterFails move 105 JsonReader.java >>>> reads int not long move 117 Timeout configuration for getAll DONE, in >>>> 0.96 119 Implement daemon functionality using Akuma skip 120 WAR >>>> dependencies are a bit of a mess skip 121 Run C++ tests in Hudson skip >>>> 128 Allow required reads to be specified during get skip, We are also >>>> thinking about changes to client api including async apis, thin client. So >>>> this will be considered in that light 130 Use variants to improve >>>> space usage of serialized VectorClock DONE 132 Cacti-formatted Output >>>> from Status Servlet skip, Not in our roadmap 140 Port >>>> RoutedStore.put() fixes to RoutedStore.delete() DONE 141 AbstractSocketPoolTest.startTest >>>> calls JUnit assert methods outside main thread move 160 Move >>>> vector-clock resolution lower in the processing stack DONE 162 Put >>>> should throw unique exceptions for its error cases (no master, too few >>>> replicas, too few successes) skip, I think the current message is >>>> descriptive 166 The enabled storage engines should be derived >>>> automatically from the stores.xml file move 169 Berkeley DB File >>>> Utilization stats should be exposed DONE 174 VoldemortConfig >>>> validation should be made uniform move 179 Study the impact of >>>> opening bdb environment per store partition Skip, With Partition based >>>> scans currently being implemented, we may not need this 180 Voldemort >>>> Consistency mechanisms Will not be executing specifically on this >>>> item. But these options will be evaluated when we get to exploring server >>>> side routing and event log feature 182 BdbStorageEngine#put may not >>>> commit or abort transaction DONE 183 AdminClient VoldemortFilter over >>>> the wire class transfer issue. skip, low priority considering filters >>>> are not very commonly used 186 truncate() failing for >>>> bdbStorageEngine. DONE 189 VersionIncrementingStore is not updating >>>> the Versioned object after a successful put skip, test only change 193 Scheduled >>>> jobs fail with no exception skip, logging issue is fixed with a big >>>> try-catch in current streaming jobs 197 Notification mechanism for >>>> changes in the cluster node membership skip 213 Add >>>> StoreClient.contains skip, There is an open pull request from roshan. >>>> Duplicate. 214 Datacenter awareness DONE 220 Add >>>> StoreClient.deleteAll(keys) skip 221 Make statistics be sliding >>>> window based instead of being reset every X seconds DONE we have >>>> implemented our histograms which does this 224 Need to differentiate >>>> between client and server timeout errors (different exceptions) DONE >>>> 225 Improved CacheStorageConfiguration skip 227 Remove calls to >>>> deprecated API in BDB4 DONE 232 voldemort binds to all interfaces >>>> instead of binding to just the host defined in the cluster.xml file for >>>> that server move 235 Performance Issues/Timeouts with NIO and Lots of >>>> Clients Skip; We use NIO and seems to be doing fine now. Also, I >>>> experimented with separation of network and disk IO in the server. What I >>>> found was that the performance actually suffered due to thread contention >>>> when disk was 'cheap' 237 Voldemort should have way of limiting the >>>> size of keys and values in a store move 238 Protocol Versioning >>>> Negotiation Needs to be Bi-directional skip, not critical 239 Voldemort >>>> should not support localhost in cluster.xml files skip 241 SocketStore >>>> redesign DONE, Jay Wylie has been working on this. If all goes well, >>>> expect some improvements in next releases 242 RoutedStore should >>>> abandon earlier if not enough nodes remain on PUT DONE all of this is >>>> already fixed/differently done in the pipeline routed store 243 Files >>>> generated by ConfigurationStorageEngine should not have xml extension >>>> unless they're actually XML skip
>>>> I will be doing the second batch in a few weeks
>> On Friday, September 21, 2012 3:42:50 PM UTC+2, Antony Stubbs wrote:
>>> Cool, but you should also change the message on the google code page >>> that says it's still used for issue tracking.
>>> On Wednesday, 19 September 2012 19:06:41 UTC-4, ctasada wrote:
>>>> Great work!
>>>> On Wed, Sep 19, 2012 at 12:30 PM, Vinoth Chandar < >>>> mail.vino...@gmail.com> wrote:
>>>>> Guys,
>>>>> I have gone over the first 48 of the tickets listed in the Google Code >>>>> at http://code.google.com/p/project-voldemort/issues/list Most of >>>>> them are fixed already or relate to non-functional aspects like code >>>>> cleanup or refer to code paths which are not the default (i.e ThreadPool >>>>> based routed store)
>>>>> I have moved over things that are either legit bug fixes or features >>>>> that fit into our roadmap. Please feel free to reopen them on github if >>>>> some issue I skipped over is very important for your use cases.
>>>>> Below is the list of tickets and rationale.
>>>>> 2 Provide publish/subscribe mechanism move; Should be addressed >>>>> by Event log; >>>>> 9 Make property names consistent move; 24 Refactoring of JDBC-based >>>>> storage engines skip; but will checkin any changes we make to MySql >>>>> engine based on storage investigation >>>>> 44 Make separate client jar skip; Non critical >>>>> 50 Json Serialization should support "not-null" feature. skip; We >>>>> are improving AVRO support. Moving towards standards 61 Finish >>>>> QueryServlet skip, Not in our roadmap >>>>> 65 MysqlGrowth uses fi instead of fj? move; 82 Run Multithreaded >>>>> tests in Hudson DONE, Hudson runs a bunch of integration like tests. >>>>> 87 Partition-Rebalance sub task 5: Partition Rebalancing User >>>>> Interface skip; Not in our roadmap 91 Implement >>>>> RoutedStoreTest.testObsoleteMasterFails move 105 JsonReader.java >>>>> reads int not long move 117 Timeout configuration for getAll DONE, >>>>> in 0.96 119 Implement daemon functionality using Akuma skip 120 WAR >>>>> dependencies are a bit of a mess skip 121 Run C++ tests in Hudson >>>>> skip 128 Allow required reads to be specified during get skip, We >>>>> are also thinking about changes to client api including async apis, thin >>>>> client. So this will be considered in that light 130 Use variants to >>>>> improve space usage of serialized VectorClock DONE 132 Cacti-formatted >>>>> Output from Status Servlet skip, Not in our roadmap 140 Port >>>>> RoutedStore.put() fixes to RoutedStore.delete() DONE 141 AbstractSocketPoolTest.startTest >>>>> calls JUnit assert methods outside main thread move 160 Move >>>>> vector-clock resolution lower in the processing stack DONE 162 Put >>>>> should throw unique exceptions for its error cases (no master, too few >>>>> replicas, too few successes) skip, I think the current message is >>>>> descriptive 166 The enabled storage engines should be derived >>>>> automatically from the stores.xml file move 169 Berkeley DB File >>>>> Utilization stats should be exposed DONE 174 VoldemortConfig >>>>> validation should be made uniform move 179 Study the impact of >>>>> opening bdb environment per store partition Skip, With Partition >>>>> based scans currently being implemented, we may not need this 180 Voldemort >>>>> Consistency mechanisms Will not be executing specifically on this >>>>> item. But these options will be evaluated when we get to exploring server >>>>> side routing and event log feature 182 BdbStorageEngine#put may not >>>>> commit or abort transaction DONE 183 AdminClient VoldemortFilter >>>>> over the wire class transfer issue. skip, low priority considering >>>>> filters are not very commonly used 186 truncate() failing for >>>>> bdbStorageEngine. DONE 189 VersionIncrementingStore is not updating >>>>> the Versioned object after a successful put skip, test only change >>>>> 193 Scheduled jobs fail with no exception skip, logging issue is >>>>> fixed with a big try-catch in current streaming jobs 197 Notification >>>>> mechanism for changes in the cluster node membership skip 213 Add >>>>> StoreClient.contains skip, There is an open pull request from roshan. >>>>> Duplicate. 214 Datacenter awareness DONE 220 Add >>>>> StoreClient.deleteAll(keys) skip 221 Make statistics be sliding >>>>> window based instead of being reset every X seconds DONE we have >>>>> implemented our histograms which does this 224 Need to differentiate >>>>> between client and server timeout errors (different exceptions) DONE >>>>> 225 Improved CacheStorageConfiguration skip 227 Remove calls to >>>>> deprecated API in BDB4 DONE 232 voldemort binds to all interfaces >>>>> instead of binding to just the host defined in the cluster.xml file for >>>>> that server move 235 Performance Issues/Timeouts with NIO and Lots >>>>> of Clients Skip; We use NIO and seems to be doing fine now. Also, I >>>>> experimented with separation of network and disk IO in the server. What I >>>>> found was that the performance actually suffered due to thread contention >>>>> when disk was 'cheap' 237 Voldemort should have way of limiting the >>>>> size of keys and values in a store move 238 Protocol Versioning >>>>> Negotiation Needs to be Bi-directional skip, not critical 239 Voldemort >>>>> should not support localhost in cluster.xml files skip 241 SocketStore >>>>> redesign DONE, Jay Wylie has been working on this. If all goes well, >>>>> expect some improvements in next releases 242 RoutedStore should >>>>> abandon earlier if not enough nodes remain on PUT DONE all of this is >>>>> already fixed/differently done in the pipeline routed store 243 Files >>>>> generated by ConfigurationStorageEngine should not have xml extension >>>>> unless they're actually XML skip
>>>>> I will be doing the second batch in a few weeks
