Specify readerID on loading history tasks #3994
Conversation
| HasNext(readerID int32) bool | ||
| Next(readerID int32) (tasks.Task, error) |
There was a problem hiding this comment.
I want iterators to be able to freely move between slices & queue readers, instead of bind to a specific slice or reader. Existing implementation is also based on that idea.
So readerID here can be regarded as a context for loading operations. The iterator implementation knows that if readerID changes, a new underlying pagingIterator is needed for talking to persistence.
yycptt
force-pushed
the
get-tasks-reader-id
branch
from
184a5cb to
723ce47
Compare
yycptt
force-pushed
the
get-tasks-reader-id
branch
from
723ce47 to
3149281
Compare
| @@ -257,6 +258,7 @@ func (r *dlqHandlerImpl) readMessagesWithAckLevel( | |||
| GetHistoryTasksRequest: persistence.GetHistoryTasksRequest{ | |||
| ShardID: r.shard.GetShardID(), | |||
| TaskCategory: tasks.CategoryReplication, | |||
| ReaderID: common.DefaultQueueReaderID, | |||
There was a problem hiding this comment.
This field probably doesn't make sense for replication DLQ.
I may later refactor the code so that GetReplicationDLQTaskRequest no longer embeds GetHistoryTasksRequest. May also do it for those dlq requests, haven't decided yet. We can sync later.
Not a problem right now cause persistence impl won't look at this field.
cc @yux0
What changed?
Why?
* The base of the PR contains History task reader persistence hint API #3986 so that review is easier, will update base after that PR is landed.How did you test it?
Potential risks
Is hotfix candidate?
no