[Tarantool-patches] [PATCH v5 2/4] replication: hide 0-th vclock components in replication responses

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Sat Apr 4 23:51:51 MSK 2020


Thanks for the patch!

Was it considered to ignore 0 component on receiver's
side rather than on sender's?

I see this:

>     This is needed for backward compatibility with old instances, which
>     don't ignore 0th vclock component coming from a remote instance by
>     default.

But can anon replicas connect to old versions?

I am not saying that it would be better, but I don't
see why technically not.

On 30/03/2020 13:04, Serge Petrenko wrote:
> If an anonymous replica is promoted to a normal one and becomes
> replication master later, its vclock contains a non-empty zero
> component, tracking local changes on this replica from the time when it
> had been anonymous. No need to pollute joining instance's vclock with
> our non-empty 0 component.
> When an anonymous replica reports its status to a remote instance it
> should also hide its 0-th vclock component.
> 
> This is needed for backward compatibility with old instances, which
> don't ignore 0th vclock component coming from a remote instance by
> default.
> Also make sure that new instances ignore 0th vclock component.
> 
> Follow-up #3186
> Prerequisite #4114
> ---
>  src/box/applier.cc             |  4 +++-
>  src/box/box.cc                 | 12 ++++++++----
>  src/box/relay.cc               |  6 ++++--
>  test/replication/anon.result   |  5 +++++
>  test/replication/anon.test.lua |  2 ++
>  5 files changed, 22 insertions(+), 7 deletions(-)
> 
> diff --git a/src/box/applier.cc b/src/box/applier.cc
> index 47a26c366..f5f67b6a9 100644
> --- a/src/box/applier.cc
> +++ b/src/box/applier.cc
> @@ -173,7 +173,9 @@ applier_writer_f(va_list ap)
>  			continue;
>  		try {
>  			struct xrow_header xrow;
> -			xrow_encode_vclock(&xrow, &replicaset.vclock);
> +			struct vclock vclock;
> +			vclock_copy_ignore0(&vclock, &replicaset.vclock);
> +			xrow_encode_vclock(&xrow, &vclock);

xrow_encode_vclock without 0 component is needed 4 times.
With 0 it is encoded 2 times. Maybe better add a function
xrow_encode_vclock_ignore0 or like that. Because copy_ignore0
is copying of ~290 bytes. This is several cache lines.

Probably even the original xrow_encode_vclock can appear to be
not needed anywhere.


More information about the Tarantool-patches mailing list