DESCRIPTION

Revert "CBQE-2340: added support to ignore expected memcached errors during load data"

This reverts commit 33aca09e0b716ae7a88b00caa52349913d853ef9.

Change-Id: Ica9bd695c841a2e04d226cf11dadda55361bfff1
Reviewed-on: http://review.couchbase.org/40494
Reviewed-by: Fisher Vargas <Fisher xxx@xxx.xxx
Tested-by: Fisher Vargas <Fisher xxx@xxx.xxx


COMMENTS

author: Fisher Vargas
date: 2014-08-11 11:07:39.644000000

Change has been successfully cherry-picked as e264bb0a1c1f3b48fe589c5808fb1bfc4f88611c

-------------------------------------
author: Esperanza Cherry
date: 2014-08-11 13:11:45.050000000

Patch Set 3:

Hi Sangharsh, when you revert something you should specify the reason, because ...

-------------------------------------
author: Fisher Vargas
date: 2014-08-11 16:18:59.467000000

Patch Set 3:

Worked on Aruna's comment last comment and figured out easy solution to this issue as of now http://review.couchbase.org/#/c/40495/ .

-------------------------------------
author: Angelina Mayer
date: 2014-08-11 18:29:04.514000000

Patch Set 3:

Hi Sangharsh,

1. If the entire commit was reverted, I'm not sure how https://www.youtube.com/watch?v=HSp9pA_o1XQ&index=3&list=PLaL-60nZw5Qel8INfF7VJDOBOGj6zZXfQ will help with new changes in _merge_keys.
If I understand correctly, the error was thrown during delete which happens well before we merge keys prior to validation. If you can also add a line explaining your solution, it will be easy to follow/review.

2. Did the test pass in your testing? If possible, if you could include a run of the test(friendpaste or the like) with the changes, it might be easier to see how the changes impact the result. If a quick check-in is needed, pls add Andrei as reviewer for changes like these. 

Thanks,
Aruna

-------------------------------------
author: Angelina Mayer
date: 2014-08-11 18:29:49.124000000

Patch Set 3:

sorry about the link, I meant http://review.couchbase.org/#/c/40495/

-------------------------------------
author: Esperanza Cherry
date: 2014-08-11 18:50:19.476000000

Patch Set 3:

btw, I think that http://review.couchbase.org/#/c/40495/ commit revert almost one to one this commit
https://github.com/couchbase/testrunner/commit/9d742a8f5de0bfc7f3c74bccef783def9188add2 

so, my thought is that after a few commits and reverts we back to what we had?

-------------------------------------
author: Angelina Mayer
date: 2014-08-11 19:12:12.156000000

Patch Set 3:

Ok looks like you've merged keys before update and deletes but I'm wondering what that will do to avoid the race between delete at source and destination. Even after merging, we are interested in source kv_store only for updates and deletes. As a result, for updates we will be updating loadOne set of keys and deleting last 30% (loadTwo) keys. That's ok but still the probability of race between two delete threads remain. Also, two merge_buckets is quite expensive.

The problem in short is this. Source delete is propagated even _before_ source KV store is updated. We should not let this happen. The solution is to have the destination delete thread wait till then.

-------------------------------------
