Cassandra write path during the decommissioning process

[Jai Bheemsen Rao Dhanwada <ja...@gmail.com>]

Hello,

I would like to understand how the writes happen when there is a node in
the cluster that is being decommissioned.

During the decommission process, the node will send the streams to the rest
of the nodes that it owns. what happened to the read and write of the data
that is owned by the node that is being decommissioned? does it even
receive reads and writes or the nodes doesn't receive any writes?

Re: Cassandra write path during the decommissioning process

[Jai Bheemsen Rao Dhanwada <ja...@gmail.com>]

Thank you

On Mon, Jun 29, 2020 at 1:01 PM Jeff Jirsa <jj...@gmail.com> wrote:

> When range movements happen (bootstrap and decom are the same here), the
> reads go to the existing replicas, and the writes go to both existing and
> the gaining replicas.
>
> This is because the decom (or bootstrap, or move) may fail at any time, so
> you need to make sure the consistency contract is maintained whether the
> range movement succeeds or fails.
>
> Quorum calculations are adjusted to accommodate this - for example, if you
> have RF=3, and you add a new host , the RF = 3 (existing) + 1 (pending),
> quorum becomes 3 of 4.
>
> On Mon, Jun 29, 2020 at 12:48 PM Jai Bheemsen Rao Dhanwada <
> jaibheemsen@gmail.com> wrote:
>
>> Can someone please share thoughts on this?
>>
>> On Wed, Jun 24, 2020 at 4:25 PM Jai Bheemsen Rao Dhanwada <
>> jaibheemsen@gmail.com> wrote:
>>
>>> Hello,
>>>
>>> I would like to understand how the writes happen when there is a node in
>>> the cluster that is being decommissioned.
>>>
>>> During the decommission process, the node will send the streams to the
>>> rest of the nodes that it owns. what happened to the read and write of the
>>> data that is owned by the node that is being decommissioned? does it even
>>> receive reads and writes or the nodes doesn't receive any writes?
>>>
>>>
>>>

Re: Cassandra write path during the decommissioning process

[Jeff Jirsa <jj...@gmail.com>]

When range movements happen (bootstrap and decom are the same here), the
reads go to the existing replicas, and the writes go to both existing and
the gaining replicas.

This is because the decom (or bootstrap, or move) may fail at any time, so
you need to make sure the consistency contract is maintained whether the
range movement succeeds or fails.

Quorum calculations are adjusted to accommodate this - for example, if you
have RF=3, and you add a new host , the RF = 3 (existing) + 1 (pending),
quorum becomes 3 of 4.

On Mon, Jun 29, 2020 at 12:48 PM Jai Bheemsen Rao Dhanwada <
jaibheemsen@gmail.com> wrote:

> Can someone please share thoughts on this?
>
> On Wed, Jun 24, 2020 at 4:25 PM Jai Bheemsen Rao Dhanwada <
> jaibheemsen@gmail.com> wrote:
>
>> Hello,
>>
>> I would like to understand how the writes happen when there is a node in
>> the cluster that is being decommissioned.
>>
>> During the decommission process, the node will send the streams to the
>> rest of the nodes that it owns. what happened to the read and write of the
>> data that is owned by the node that is being decommissioned? does it even
>> receive reads and writes or the nodes doesn't receive any writes?
>>
>>
>>

Re: Cassandra write path during the decommissioning process

[Jai Bheemsen Rao Dhanwada <ja...@gmail.com>]

Can someone please share thoughts on this?

On Wed, Jun 24, 2020 at 4:25 PM Jai Bheemsen Rao Dhanwada <
jaibheemsen@gmail.com> wrote:

> Hello,
>
> I would like to understand how the writes happen when there is a node in
> the cluster that is being decommissioned.
>
> During the decommission process, the node will send the streams to the
> rest of the nodes that it owns. what happened to the read and write of the
> data that is owned by the node that is being decommissioned? does it even
> receive reads and writes or the nodes doesn't receive any writes?
>
>
>