You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cassandra.apache.org by Benjamin Lerer <b....@gmail.com> on 2021/11/29 16:58:34 UTC
[DISCUSS] Throughput issues when inserting on contended partitions
Hi everybody,
We have seen some serious throughput issues when inserting data with
collections on contended partitions ( CASSANDRA-17163
<https://issues.apache.org/jira/browse/CASSANDRA-17163> and CASSANDRA-15464
<https://issues.apache.org/jira/browse/CASSANDRA-15464>).
Benedict has created some patches to address those issues and improve the
insertion throughput and memory consumptions ( CASSANDRA-15510
<https://issues.apache.org/jira/browse/CASSANDRA-15510> and CASSANDRA-15511
<https://issues.apache.org/jira/browse/CASSANDRA-15511>)
Those patches are significant changes and are currently marked as
improvements.
I am wondering if we should not consider implementing those changes in 4.0
considering the fact that they will fix serious existing issues and would
like to hear your options about it.
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by Benjamin Lerer <b....@gmail.com>.
Thanks for your answer. 😀
I will review the patches and port them to 4.0
Le jeu. 2 déc. 2021 à 07:08, Berenguer Blasi <be...@gmail.com> a
écrit :
> +1 it's severe enough I also think it should go in.
>
> On 1/12/21 18:10, Dinesh Joshi wrote:
> > +1
> >
> > Dinesh
> >
> >> On Dec 1, 2021, at 5:59 AM, Mick Semb Wever <mc...@apache.org> wrote:
> >>
> >> 
> >>> I am wondering if we should not consider implementing those changes in
> 4.0
> >>> considering the fact that they will fix serious existing issues and
> would
> >>> like to hear your options about it.
> >>
> >> +1 to see them land in a 4.0.x
> >>
> >> This has hit a number of users, and is an unfortunate reason some go
> >> back to using `memtable_allocation_type: heap_buffers` (which isn't a
> >> fix for the problem and only barely improves their situation).
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> >> For additional commands, e-mail: dev-help@cassandra.apache.org
> >>
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> > For additional commands, e-mail: dev-help@cassandra.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> For additional commands, e-mail: dev-help@cassandra.apache.org
>
>
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by Berenguer Blasi <be...@gmail.com>.
+1 it's severe enough I also think it should go in.
On 1/12/21 18:10, Dinesh Joshi wrote:
> +1
>
> Dinesh
>
>> On Dec 1, 2021, at 5:59 AM, Mick Semb Wever <mc...@apache.org> wrote:
>>
>> 
>>> I am wondering if we should not consider implementing those changes in 4.0
>>> considering the fact that they will fix serious existing issues and would
>>> like to hear your options about it.
>>
>> +1 to see them land in a 4.0.x
>>
>> This has hit a number of users, and is an unfortunate reason some go
>> back to using `memtable_allocation_type: heap_buffers` (which isn't a
>> fix for the problem and only barely improves their situation).
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
>> For additional commands, e-mail: dev-help@cassandra.apache.org
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> For additional commands, e-mail: dev-help@cassandra.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by Dinesh Joshi <dj...@apache.org>.
+1
Dinesh
> On Dec 1, 2021, at 5:59 AM, Mick Semb Wever <mc...@apache.org> wrote:
>
> 
>>
>> I am wondering if we should not consider implementing those changes in 4.0
>> considering the fact that they will fix serious existing issues and would
>> like to hear your options about it.
>
>
> +1 to see them land in a 4.0.x
>
> This has hit a number of users, and is an unfortunate reason some go
> back to using `memtable_allocation_type: heap_buffers` (which isn't a
> fix for the problem and only barely improves their situation).
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> For additional commands, e-mail: dev-help@cassandra.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by Mick Semb Wever <mc...@apache.org>.
> I am wondering if we should not consider implementing those changes in 4.0
> considering the fact that they will fix serious existing issues and would
> like to hear your options about it.
+1 to see them land in a 4.0.x
This has hit a number of users, and is an unfortunate reason some go
back to using `memtable_allocation_type: heap_buffers` (which isn't a
fix for the problem and only barely improves their situation).
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by "benedict@apache.org" <be...@apache.org>.
I’m in favour, though I have weaker requirements for backports than others.
This work is pretty significant, though. It’s nothing like the complexity of CEP-14, but it heavily modifies a critical piece of the system. I would say that it needs a rigorous review process if it’s going into a patch release.
I also won’t have time to forward port it in the near future, though I think it shouldn’t be too onerous for other contributors, and I’m happy to help guide that process.
From: Brandon Williams <dr...@gmail.com>
Date: Monday, 29 November 2021 at 17:03
To: dev@cassandra.apache.org <de...@cassandra.apache.org>
Subject: Re: [DISCUSS] Throughput issues when inserting on contended partitions
I think it makes a lot of sense to fix these in 4.0, they have been
lingering issues long enough. +1
On Mon, Nov 29, 2021 at 10:59 AM Benjamin Lerer <b....@gmail.com> wrote:
>
> Hi everybody,
>
> We have seen some serious throughput issues when inserting data with
> collections on contended partitions ( CASSANDRA-17163
> <https://issues.apache.org/jira/browse/CASSANDRA-17163> and CASSANDRA-15464
> <https://issues.apache.org/jira/browse/CASSANDRA-15464>).
> Benedict has created some patches to address those issues and improve the
> insertion throughput and memory consumptions ( CASSANDRA-15510
> <https://issues.apache.org/jira/browse/CASSANDRA-15510> and CASSANDRA-15511
> <https://issues.apache.org/jira/browse/CASSANDRA-15511>)
> Those patches are significant changes and are currently marked as
> improvements.
>
> I am wondering if we should not consider implementing those changes in 4.0
> considering the fact that they will fix serious existing issues and would
> like to hear your options about it.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org
Re: [DISCUSS] Throughput issues when inserting on contended partitions
Posted by Brandon Williams <dr...@gmail.com>.
I think it makes a lot of sense to fix these in 4.0, they have been
lingering issues long enough. +1
On Mon, Nov 29, 2021 at 10:59 AM Benjamin Lerer <b....@gmail.com> wrote:
>
> Hi everybody,
>
> We have seen some serious throughput issues when inserting data with
> collections on contended partitions ( CASSANDRA-17163
> <https://issues.apache.org/jira/browse/CASSANDRA-17163> and CASSANDRA-15464
> <https://issues.apache.org/jira/browse/CASSANDRA-15464>).
> Benedict has created some patches to address those issues and improve the
> insertion throughput and memory consumptions ( CASSANDRA-15510
> <https://issues.apache.org/jira/browse/CASSANDRA-15510> and CASSANDRA-15511
> <https://issues.apache.org/jira/browse/CASSANDRA-15511>)
> Those patches are significant changes and are currently marked as
> improvements.
>
> I am wondering if we should not consider implementing those changes in 4.0
> considering the fact that they will fix serious existing issues and would
> like to hear your options about it.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org