You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@falcon.apache.org by Suhas Vasu <su...@gmail.com> on 2015/02/02 13:51:39 UTC

Re: Review Request 30116: Parallel update APIs create 2 coords

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/
-----------------------------------------------------------

(Updated Feb. 2, 2015, 12:51 p.m.)


Review request for Falcon.


Changes
-------

Updated patch based on review comments


Repository: falcon-git


Description
-------

In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
This leads to multiple coordinators being spawned in Oozie.

We need to ensure that the update method is syncronized.


Diffs (updated)
-----

  common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
  prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 

Diff: https://reviews.apache.org/r/30116/diff/


Testing
-------

UT's were added and were successful


Thanks,

Suhas  Vasu


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Shwetha GS <ss...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review72579
-----------------------------------------------------------

Ship it!


Minor comments. Looks good otherwise


prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118635>

    entity.toShortString() already contain entity name



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118640>

    e.toShortString() already contain name



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118641>

    Use shortString()


- Shwetha GS


On Feb. 13, 2015, 1:54 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 13, 2015, 1:54 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   common/src/test/java/org/apache/falcon/entity/lock/MemoryLocksTest.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
>   prism/src/main/java/org/apache/falcon/resource/AbstractSchedulableEntityManager.java b8d12ee 
>   webapp/src/test/java/org/apache/falcon/cli/FalconCLIIT.java 118003f 
>   webapp/src/test/java/org/apache/falcon/resource/EntityManagerJerseyIT.java 40f8e04 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Suhas Vasu <su...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/
-----------------------------------------------------------

(Updated Feb. 13, 2015, 1:54 p.m.)


Review request for Falcon.


Changes
-------

Fixing minor nits as per review comments


Repository: falcon-git


Description
-------

In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
This leads to multiple coordinators being spawned in Oozie.

We need to ensure that the update method is syncronized.


Diffs (updated)
-----

  common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
  common/src/test/java/org/apache/falcon/entity/lock/MemoryLocksTest.java PRE-CREATION 
  prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
  prism/src/main/java/org/apache/falcon/resource/AbstractSchedulableEntityManager.java b8d12ee 
  webapp/src/test/java/org/apache/falcon/cli/FalconCLIIT.java 118003f 
  webapp/src/test/java/org/apache/falcon/resource/EntityManagerJerseyIT.java 40f8e04 

Diff: https://reviews.apache.org/r/30116/diff/


Testing
-------

UT's were added and were successful


Thanks,

Suhas  Vasu


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Shwetha GS <ss...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review72133
-----------------------------------------------------------



common/src/test/resources/config/process/process-duplicate-name.xml
<https://reviews.apache.org/r/30116/#comment118147>

    Do you need another test xml for this? Just set process name after unmarshal?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118148>

    also include entity type. You can use entity.toShortString()



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118149>

    why release here? Won't finally take care of it?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118150>

    not in submit, schedule right?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment118151>

    why catch Throwable?



webapp/src/test/java/org/apache/falcon/resource/EntityManagerJerseyIT.java
<https://reviews.apache.org/r/30116/#comment118153>

    Add Thread.join()


- Shwetha GS


On Feb. 11, 2015, 10:15 a.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 11, 2015, 10:15 a.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   common/src/test/java/org/apache/falcon/entity/lock/MemoryLocksTest.java PRE-CREATION 
>   common/src/test/resources/config/process/process-duplicate-name.xml PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
>   webapp/src/test/java/org/apache/falcon/cli/FalconCLIIT.java 118003f 
>   webapp/src/test/java/org/apache/falcon/resource/EntityManagerJerseyIT.java 40f8e04 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Suhas Vasu <su...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/
-----------------------------------------------------------

(Updated Feb. 11, 2015, 10:15 a.m.)


Review request for Falcon.


Changes
-------

Updating the patch based on review comments


Repository: falcon-git


Description
-------

In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
This leads to multiple coordinators being spawned in Oozie.

We need to ensure that the update method is syncronized.


Diffs (updated)
-----

  common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
  common/src/test/java/org/apache/falcon/entity/lock/MemoryLocksTest.java PRE-CREATION 
  common/src/test/resources/config/process/process-duplicate-name.xml PRE-CREATION 
  prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
  webapp/src/test/java/org/apache/falcon/cli/FalconCLIIT.java 118003f 
  webapp/src/test/java/org/apache/falcon/resource/EntityManagerJerseyIT.java 40f8e04 

Diff: https://reviews.apache.org/r/30116/diff/


Testing
-------

UT's were added and were successful


Thanks,

Suhas  Vasu


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Pallavi Rao <pa...@inmobi.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review70704
-----------------------------------------------------------

Ship it!


Ship It!

- Pallavi Rao


On Feb. 2, 2015, 12:51 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2015, 12:51 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Suhas Vasu <su...@gmail.com>.

> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 57
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line57>
> >
> >     this should be synchronised so that 2 threads will not get the same lock
> 
> Suhas  Vasu wrote:
>     I am just relying on the concurrent hashmap to handle the syncronization of adding to the map. should be ok right ?
> 
> Shwetha GS wrote:
>     Take an example of 2 threads trying to acquire lock for the same entity at the same time with this sequence of execution(assuming its not already locked):
>     t1 -> !locks.containsKey(entityName) will return true
>     t2 -> !locks.containsKey(entityName) will also return true
>     t1 -> locks.put(entityName, MAP_VALUE_CONSTANT) will get lock
>     t2 -> locks.put(entityName, MAP_VALUE_CONSTANT) will also get lock
>     
>     Since acquireLock() is not synchronised, both t1 and t2 will get the lock in this case.
>     
>     Instead of making the method synchronised, you can use ConcurrentHashMap.putIfAbsent() and check the return type if lock is successful. putIfAbsent is atomic

oh, I missed this scenario


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 334
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line334>
> >
> >     acquireLock() doesn't throw exception?
> 
> Suhas  Vasu wrote:
>     It might. In acquire lock we add the entry into a concurrent hash map. It might throw up in case of concurrent writes.
> 
> Shwetha GS wrote:
>     acquireLock() signature doesn't define any exception. When is the exception thrown?

Have removed this as the signature of the underlying method has changed.


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 354
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line354>
> >
> >     Update API actually takes longer because of dryRun that you should be able to add IT. Can you check please?
> 
> Suhas  Vasu wrote:
>     I tried, didn't work. will crosscheck once

The issue is trying to simulate 2 independent & simultaneous update commands. The code gets too complicated while trying to achieve this.


- Suhas


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review70728
-----------------------------------------------------------


On Feb. 2, 2015, 12:51 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2015, 12:51 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Shwetha GS <ss...@hortonworks.com>.

> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 42
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line42>
> >
> >     you can instantiate this at 'private static MemoryLocks instance' to avoid getInstance() synchronised
> 
> Suhas  Vasu wrote:
>     Out of curiousity, advantages of changing this to private static over synchronized getInstance() ?

getInstance() is called from a lot of places. Instead of synchronising each call, initialising at static variable(which is guaranteed to be called once) is better.


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 57
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line57>
> >
> >     this should be synchronised so that 2 threads will not get the same lock
> 
> Suhas  Vasu wrote:
>     I am just relying on the concurrent hashmap to handle the syncronization of adding to the map. should be ok right ?

Take an example of 2 threads trying to acquire lock for the same entity at the same time with this sequence of execution(assuming its not already locked):
t1 -> !locks.containsKey(entityName) will return true
t2 -> !locks.containsKey(entityName) will also return true
t1 -> locks.put(entityName, MAP_VALUE_CONSTANT) will get lock
t2 -> locks.put(entityName, MAP_VALUE_CONSTANT) will also get lock

Since acquireLock() is not synchronised, both t1 and t2 will get the lock in this case.

Instead of making the method synchronised, you can use ConcurrentHashMap.putIfAbsent() and check the return type if lock is successful. putIfAbsent is atomic


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 334
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line334>
> >
> >     acquireLock() doesn't throw exception?
> 
> Suhas  Vasu wrote:
>     It might. In acquire lock we add the entry into a concurrent hash map. It might throw up in case of concurrent writes.

acquireLock() signature doesn't define any exception. When is the exception thrown?


- Shwetha


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review70728
-----------------------------------------------------------


On Feb. 2, 2015, 12:51 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2015, 12:51 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Suhas Vasu <su...@gmail.com>.

> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 32
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line32>
> >
> >     Use some small data like Bool(so that memory used is less) as this is constant and not used anywhere

makes sense, will do


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 42
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line42>
> >
> >     you can instantiate this at 'private static MemoryLocks instance' to avoid getInstance() synchronised

Out of curiousity, advantages of changing this to private static over synchronized getInstance() ?


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 57
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line57>
> >
> >     this should be synchronised so that 2 threads will not get the same lock

I am just relying on the concurrent hashmap to handle the syncronization of adding to the map. should be ok right ?


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java, line 81
> > <https://reviews.apache.org/r/30116/diff/2/?file=843611#file843611line81>
> >
> >     Add UTs for Memory Locks

Will do


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 282
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line282>
> >
> >     tokenList can never be empty?

Yes, the token list will not be empty. Its redundant.


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 334
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line334>
> >
> >     acquireLock() doesn't throw exception?

It might. In acquire lock we add the entry into a concurrent hash map. It might throw up in case of concurrent writes.


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 344
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line344>
> >
> >     we should get just entity lock(not dependents) at schedule() as well

Makes sense


> On Feb. 3, 2015, 10:27 a.m., Shwetha GS wrote:
> > prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java, line 354
> > <https://reviews.apache.org/r/30116/diff/2/?file=843612#file843612line354>
> >
> >     Update API actually takes longer because of dryRun that you should be able to add IT. Can you check please?

I tried, didn't work. will crosscheck once


- Suhas


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review70728
-----------------------------------------------------------


On Feb. 2, 2015, 12:51 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2015, 12:51 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>


Re: Review Request 30116: Parallel update APIs create 2 coords

Posted by Shwetha GS <ss...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30116/#review70728
-----------------------------------------------------------



common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java
<https://reviews.apache.org/r/30116/#comment116099>

    Use some small data like Bool(so that memory used is less) as this is constant and not used anywhere



common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java
<https://reviews.apache.org/r/30116/#comment116100>

    you can instantiate this at 'private static MemoryLocks instance' to avoid getInstance() synchronised



common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java
<https://reviews.apache.org/r/30116/#comment116102>

    this should be synchronised so that 2 threads will not get the same lock



common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java
<https://reviews.apache.org/r/30116/#comment116108>

    Add UTs for Memory Locks



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment116103>

    static?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment116104>

    tokenList can never be empty?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment116109>

    acquireLock() doesn't throw exception?



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment116110>

    we should get just entity lock(not dependents) at schedule() as well



prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java
<https://reviews.apache.org/r/30116/#comment116111>

    Update API actually takes longer because of dryRun that you should be able to add IT. Can you check please?


- Shwetha GS


On Feb. 2, 2015, 12:51 p.m., Suhas  Vasu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30116/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2015, 12:51 p.m.)
> 
> 
> Review request for Falcon.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> In distributed mode, when parallel update command are issued for same entity, they are not syncronized.
> This leads to multiple coordinators being spawned in Oozie.
> 
> We need to ensure that the update method is syncronized.
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/entity/lock/MemoryLocks.java PRE-CREATION 
>   prism/src/main/java/org/apache/falcon/resource/AbstractEntityManager.java 0d34ef3 
> 
> Diff: https://reviews.apache.org/r/30116/diff/
> 
> 
> Testing
> -------
> 
> UT's were added and were successful
> 
> 
> Thanks,
> 
> Suhas  Vasu
> 
>