Postgres Pro
Downloads
Home   >   mailing lists

Thread: [HACKERS] GSoC 2017 weekly progress reports (week 2)

[HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Shubham Barai
Date:
Project: Explicitly support predicate locks in index AMs besides b-tree

Hi, 

During this week, I mostly worked on testing to verify my code and on debugging to solve some issues I was having. I have specifically created two tests. The first test is about verifying serialization failure when there is a read-write conflict. The second one is about checking false positives. We need to make sure that it doesn't generate false positive serialization failures.  So far, I have got expected results. Any feedback on results will be appreciated.


For more details about testing, please have a look at attached files.




Regards,
Shubham



Sent with Mailtrack
Attachment

Re: [HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Andrew Borodin
Date:
2017-06-13 18:00 GMT+05:00 Shubham Barai <shubhambaraiss@gmail.com>:
>
> Project: Explicitly support predicate locks in index AMs besides b-tree
>
Hi, Shubham
Good job!

So, in current HEAD test predicate_gist_2.spec generate false
positives, but with your patch, it does not?
I'd suggest keeping spec tests with your code in the same branch, it's
easier. Also it worth to clean up specs style and add some words to
documentation.

Kevin, all, how do you think, is it necessary to expand these tests
not only on Index Only Scan, but also on Bitmap Index Scan? And may be
KNN version of scan too?
I couldn't find such tests for B-tree, do we have them?

Best regards, Andrey Borodin, Octonica.

Re: [HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Kevin Grittner
Date:
On Tue, Jun 13, 2017 at 1:02 PM, Andrew Borodin <borodin@octonica.com> wrote:
> 2017-06-13 18:00 GMT+05:00 Shubham Barai <shubhambaraiss@gmail.com>:

> Good job!

+1!  :-)

> So, in current HEAD test predicate_gist_2.spec generate false
> positives, but with your patch, it does not?

Keep in mind, that false positives do not break *correctness* of serializable
transactions as long as it involves another transaction.  (It *would* be a bug
if a transaction running alone could cause a serialization failure.)  A false
*negative* is always a bug.

That said, false positives hurt performance, so we should keep the rate as low
as practicable.

> I'd suggest keeping spec tests with your code in the same branch, it's
> easier.

+1

> Also it worth to clean up specs style and add some words to
> documentation.

+1

> Kevin, all, how do you think, is it necessary to expand these tests
> not only on Index Only Scan, but also on Bitmap Index Scan? And may be
> KNN version of scan too?
> I couldn't find such tests for B-tree, do we have them?

Off-hand, I don't know.  It would be interesting to run regression tests with
code coverage and look at the index AMs.

https://www.postgresql.org/docs/9.6/static/regress-coverage.html

-- 
Kevin Grittner
VMware vCenter Server
https://www.vmware.com/

Re: [HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Shubham Barai
Date:
Hi,

I have made some changes in tests and pushed them to my branch.

Thanks for helping me out with testing.

Now, current head produces false positives but, with my patch, it doesn't.

Here is the link for updated tests: https://github.com/shubhambaraiss/postgres/commit/2c02685a50a2b30654beb5c52542a57a46219c39

 

Regards,
Shubham



Sent with Mailtrack

On 13 June 2017 at 23:32, Andrew Borodin <borodin@octonica.com> wrote:
2017-06-13 18:00 GMT+05:00 Shubham Barai <shubhambaraiss@gmail.com>:
>
> Project: Explicitly support predicate locks in index AMs besides b-tree
>
Hi, Shubham
Good job!

So, in current HEAD test predicate_gist_2.spec generate false
positives, but with your patch, it does not?
I'd suggest keeping spec tests with your code in the same branch, it's
easier. Also it worth to clean up specs style and add some words to
documentation.

Kevin, all, how do you think, is it necessary to expand these tests
not only on Index Only Scan, but also on Bitmap Index Scan? And may be
KNN version of scan too?
I couldn't find such tests for B-tree, do we have them?

Best regards, Andrey Borodin, Octonica.

Re: [HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Alvaro Herrera
Date:
Shubham Barai wrote:
> Hi,
> 
> I have made some changes in tests and pushed them to my branch.
> 
> Thanks for helping me out with testing.
> 
> Now, current head produces false positives but, with my patch, it doesn't.
> 
> Here is the link for updated tests:
> https://github.com/shubhambaraiss/postgres/commit/2c02685a50a2b30654beb5c52542a57a46219c39

Nice.  Please provide patches, so that it can be considered for commit.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Re: [HACKERS] GSoC 2017 weekly progress reports (week 2)

From
Shubham Barai
Date:


On 15 June 2017 at 07:23, Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
Shubham Barai wrote:
> Hi,
>
> I have made some changes in tests and pushed them to my branch.
>
> Thanks for helping me out with testing.
>
> Now, current head produces false positives but, with my patch, it doesn't.
>
> Here is the link for updated tests:
> https://github.com/shubhambaraiss/postgres/commit/2c02685a50a2b30654beb5c52542a57a46219c39

Nice.  Please provide patches, so that it can be considered for commit.
 
Yes, I will provide a patch shortly. I just need to update documentation first. 

--
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services




Sent with Mailtrack