Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r46264229
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateBackend.java
---
@@ -0,0 +1,248 @@
+/*
Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r46264128
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateBackend.java
---
@@ -0,0 +1,248 @@
+/*
Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r46264109
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateBackend.java
---
@@ -0,0 +1,248 @@
+/*
Github user asfgit closed the pull request at:
https://github.com/apache/flink/pull/1305
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enab
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-159205217
I think you can go ahead. It's in contrib and you guys are battle-testing
it anyways... :wink:
---
If your project is set up for it, you can reply to this email and h
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-159048039
If no objections I would like to merge this :)
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your pro
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-158799081
@StephanEwen, @rmetzger:
I addressed the comments regarding the logs and the state id.
I also added a final improvement:
-Now compaction is executed i
Github user rmetzger commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45544375
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateHandle.java
---
@@ -0,0 +1,88 @@
+/*
+
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45479951
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateHandle.java
---
@@ -0,0 +1,88 @@
+/*
+ *
Github user rmetzger commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45479621
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbStateHandle.java
---
@@ -0,0 +1,88 @@
+/*
+
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-158423218
Had an offline chat with @gyfora with the following outcome:
- A deterministic state identifier is needed
- Small change to pass that identifier as a singl
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45474858
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state bac
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45473713
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state backend,
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45377370
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state backend,
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45373479
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state bac
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45372569
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state backend,
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45372620
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state bac
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45372303
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state backend,
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45371919
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state bac
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45371097
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state backend,
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-158107515
I have a final comment inline. Otherwise, I think this is good to merge.
---
If your project is set up for it, you can reply to this email and have your
reply appear
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45362408
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/state/StateBackend.java ---
@@ -64,33 +64,35 @@
/**
* Closes the state bac
Github user gyfora commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45351896
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbBackendConfig.java
---
@@ -0,0 +1,406 @@
+/*
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/1305#discussion_r45351298
--- Diff:
flink-contrib/flink-streaming-contrib/src/main/java/org/apache/flink/contrib/streaming/state/DbBackendConfig.java
---
@@ -0,0 +1,406 @@
+/*
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-158074782
Looking though this again...
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156964684
I'm looking at it again.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156809400
I have updated the description, and ran some more cluster tests without any
issues.
It would be good if you all could do a second round of reviews please.
---
I
Github user uce commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156064645
I totally understand your point, but I think it's OK that changes of this
scope take longer to review and get in (my HA PR took over a month or so to get
in). At the end of
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156039726
Well, I don't know what they are working on... It would be easier not
having to rebase state backend api changes
---
If your project is set up for it, you can reply to t
Github user uce commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156038439
I agree that this will be an important backend and good to have in. :) But
do we need to push this right now? I think we should wait a little and make
sure that it fits well
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-156034877
I would like to push this soon if no objections
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your pr
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-155765000
Should we do a final iteration over this and merge this to contrib?
The description got slightly out of date when I changed this back so that
it stores the state
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153872782
I updated the sharding logic to do mod hashing by default on the keys for
the number of shards, and the user can also add a custom Partitioner to
implement custom shardin
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153696208
I also removed the sharding logic now as I think it was pretty weak and not
very useful (it maintained 1 connection per subtask which would break if we
change parallelism
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153695943
I updated this PR with the reworked logic, it has several advantages over
the previous timestamp based solution (including the elimination of
transactions from the logic)
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153500173
I am now writing a prototype for this version, the batch insert got pretty
ugly...
I will probably finish it tomorrow.
---
If your project is set up for it, you can
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153470424
After some initial discussion we @StephanEwen we came to the following
conclusions:
The current timestamp based approach has some limitations in terms of the
ass
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153376152
My last commit introduces automatic compaction with user specified
frequency. It also allows the KvStates to implement the CheckpointNotifier
interface in which case they
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153351332
@gyfora That's what I meant, basically the timestamps could subsume the
role of the checkpointIds. I.e. The checkpointIds have the semantics of the
timestamps and the t
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153339094
@StephanEwen
Thanks for the comments. You are right the main idea is exactly as you
described.
The reason why exactly-once is violated in some corner cases becau
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153337412
You are right about the checkpointIDs being ignored. I don't see why we
need to bother with changing the id semantics, the timestamps also serve as
checkpoint Ids perfect
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-15538
2. Ah, I meant the lookupTimestamp. In an earlier version you used both the
checkpointId and lookupTimestamp to perform key lookups.
3. I see, in this implementation
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153307495
@aljoscha
1. I was initially using the MockEnvironments but I added the
DummyEnvironment for several reasons: I wanted control over the JobId and the
number of subt
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153305943
Just some remarks:
- `DummyEnvironment` seems unnecessary, we already have
`StreamMockEnvironment`. I think it could be reused.
- In the first version you
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-153289042
Cool stuff, really! This is very much in line with what I had in mind for a
SQL backend.
Let me check if I understood everything correct (and see where my
u
Github user uce commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-152948336
Thanks for the great write up!
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-152944197
Good stuff! Will need a day more to look through this, but this is a cool
way of doing stateful stream computation :-)
---
If your project is set up for it, you can
Github user gyfora commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-152767697
My last commit (that was meant to solve the problems with failed tasks
writing to the db) introduced some issues with the exactly once guarantees. I
will look into it tom
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/1305#issuecomment-151884357
Wow, a lot of stuff. I will look into it once the release is out. :smiley:
---
If your project is set up for it, you can reply to this email and have your
reply appear
GitHub user gyfora opened a pull request:
https://github.com/apache/flink/pull/1305
Out-of-core state backend for JDBC databases
Detailed description incoming...
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/gyfora/flink master
50 matches
Mail list logo