> On July 31, 2014, 11:13 p.m., Bill Farner wrote: > > src/main/thrift/org/apache/aurora/gen/api.thrift, line 546 > > <https://reviews.apache.org/r/24116/diff/3/?file=647049#file647049line546> > > > > I find the latest revision of the structs a bit tough to comprehend. > > How about this tweak: > > > > struct JobUpdateRequest { > > // fields currently in UpdateSettings and UpdateConfiguration > > } > > > > // the high-level view of what an update intended to do, and its > > current status > > struct JobUpdate { > > 1: string id > > 2: JobKey job > > 3: TaskConfig config > > 4: i32 instances > > 5: UpdateStatus status > > 6: i64 startTimestampMs > > 7: i64 endTimestampMs > > } > > > > enum InstanceUpdateAction { > > ROLLED_FORWARD, > > ROLLED_BACK, > > ADDED, > > REMOVED > > } > > > > // the individual actions taken as part of an update > > struct InstanceUpdateEvent { > > 1: i32 instance > > 2: InstanceUpdateAction action > > 3: i64 timestampMs > > } > > > > These two structs would be fetched via two separate API calls. > > Maxim Khutornenko wrote: > How does JobUpdate fit with the getUpdates() API? Do you propose we > return JobUpdate + map<i32, InstanceUpdateEvent> with it? > > I am not sure I like the fact that we are mixing immutable and mutable > data within the same return struct. This would mean we always return > TaskConfig even when it's not really needed. > > I'd rather go with Event-only getUpdates() API and have something like > getUpdateDetails(string id) that would return JobUpdate with immutable-only > data. This approach is more like the current diff + a new API to return > JobUpdate only.
In the above arrangement, getUpdates() would return set<JobUpdate>, another API would return the events associated with a specific update. (That's what i meant by "These two structs would be fetched via two separate API calls.") - Bill ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/24116/#review49301 ----------------------------------------------------------- On July 31, 2014, 6:37 p.m., Maxim Khutornenko wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/24116/ > ----------------------------------------------------------- > > (Updated July 31, 2014, 6:37 p.m.) > > > Review request for Aurora, David McLaughlin and Bill Farner. > > > Bugs: AURORA-611 > https://issues.apache.org/jira/browse/AURORA-611 > > > Repository: aurora > > > Description > ------- > > First stab at update APIs. > > > Diffs > ----- > > > src/main/java/org/apache/aurora/scheduler/thrift/SchedulerThriftInterface.java > 12de5a3e9e3aae217b30c385e2d7ec7b43863ae2 > src/main/thrift/org/apache/aurora/gen/api.thrift > 54b8985971719247a5d42d8676075a51045bbb92 > src/test/java/org/apache/aurora/scheduler/thrift/aop/ForwardingThrift.java > 2ea4a9ba0a1ea81fea5c4f5203457aa79ae67c10 > > Diff: https://reviews.apache.org/r/24116/diff/ > > > Testing > ------- > > gradle build > > > Thanks, > > Maxim Khutornenko > >
