Realtime get from in-memory segment when possible#64504
Merged
dnhatn merged 14 commits intoelastic:masterfrom Nov 10, 2020
Merged
Realtime get from in-memory segment when possible#64504dnhatn merged 14 commits intoelastic:masterfrom
dnhatn merged 14 commits intoelastic:masterfrom
Conversation
Collaborator
|
Pinging @elastic/es-distributed (:Distributed/Engine) |
Contributor
henningandersen
left a comment
There was a problem hiding this comment.
Thanks @dnhatn , this looks good. I have only looked at the main part of the PR for now and have a single comment/question.
server/src/main/java/org/elasticsearch/index/engine/InternalEngine.java
Outdated
Show resolved
Hide resolved
Contributor
henningandersen
left a comment
There was a problem hiding this comment.
A few smaller comments, otherwise looking good.
server/src/main/java/org/elasticsearch/index/engine/SingleDocDirectoryReader.java
Outdated
Show resolved
Hide resolved
|
|
||
| @Override | ||
| public CacheHelper getReaderCacheHelper() { | ||
| return in.getReaderCacheHelper(); |
Contributor
There was a problem hiding this comment.
Might mostly be for my understanding, but since the wrapped directory reader could return different results from the inner reader, should we not return null here?
Member
Author
There was a problem hiding this comment.
Good catch. We should return null here as we modify the liveDocs. I addressed it in eb968e4.
Member
Author
There was a problem hiding this comment.
I had to reverted this and added a comment instead. Please see: e176bf2
jpountz
approved these changes
Nov 4, 2020
server/src/main/java/org/elasticsearch/index/engine/Engine.java
Outdated
Show resolved
Hide resolved
server/src/main/java/org/elasticsearch/index/engine/Engine.java
Outdated
Show resolved
Hide resolved
server/src/main/java/org/elasticsearch/index/engine/SingleDocDirectoryReader.java
Outdated
Show resolved
Hide resolved
Member
Author
|
@henningandersen Are you okay with the latest changes? |
henningandersen
approved these changes
Nov 9, 2020
server/src/test/java/org/elasticsearch/index/engine/InternalEngineTests.java
Show resolved
Hide resolved
Member
Author
|
@henningandersen @jpountz Thanks for your reviews. |
dnhatn
added a commit
that referenced
this pull request
Nov 10, 2020
If the reader wrapper is specified, then we can't perform a realtime get using operations from translog. With this change, we will create an in-memory Lucene segment from that indexing operation and perform a realtime get from that segment to avoid refresh storms.
dnhatn
added a commit
that referenced
this pull request
Nov 10, 2020
ywelsch
added a commit
that referenced
this pull request
Jul 1, 2021
Reading from translog during a realtime get requires special handling in some higher level components, e.g. ShardGetService, where we're doing a bunch of tricks to extract other stored fields from the source. Another issue with the current approach relates to #74227 where we introduce a new "field usage tracking" directory wrapper that's always applied, and we want to make sure that we can still quickly do realtime gets from translog without creating an in-memory index of the document, even when this directory wrapper exists. This PR introduces a directory reader that contains a single translog indexing operation. This can be used during a realtime get to access documents that haven't been refreshed yet. In the normal case, all information relevant to resolve the realtime get is mocked out to provide fast access to _id and _source. In case where more values are requested (e.g. access to other stored fields) etc., this reader will index the document into an in-memory Lucene segment that is created on-demand. Relates #64504
ywelsch
added a commit
that referenced
this pull request
Jul 1, 2021
Reading from translog during a realtime get requires special handling in some higher level components, e.g. ShardGetService, where we're doing a bunch of tricks to extract other stored fields from the source. Another issue with the current approach relates to #74227 where we introduce a new "field usage tracking" directory wrapper that's always applied, and we want to make sure that we can still quickly do realtime gets from translog without creating an in-memory index of the document, even when this directory wrapper exists. This PR introduces a directory reader that contains a single translog indexing operation. This can be used during a realtime get to access documents that haven't been refreshed yet. In the normal case, all information relevant to resolve the realtime get is mocked out to provide fast access to _id and _source. In case where more values are requested (e.g. access to other stored fields) etc., this reader will index the document into an in-memory Lucene segment that is created on-demand. Relates #64504
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
If the reader wrapper is specified, then we can't perform a realtime get using operations from translog. With this change, we will create an in-memory Lucene segment from that indexing operation and perform a realtime get from that segment to avoid refresh storms.