[Tarantool-patches] [PATCH 7/7] box: make promote/demote always bump the term

[Vladislav Shpilevoy]

Thanks for the patch!

See 4 comments below.

>     box: make promote always bump the term
>     
>     When called without elections, promote and resulted in multiple

1. 'and' is excess.

>     PROMOTE entries for the same term. This is not right, because all
>     the promotions for the same term except the first one would be ignored
>     as already seen.
>     
>     Part-of #6034
> 
> diff --git a/src/box/box.cc b/src/box/box.cc
> index 6a0950f44..53a8f80e5 100644
> --- a/src/box/box.cc
> +++ b/src/box/box.cc
> @@ -1687,16 +1687,18 @@ box_promote(void)
>  			rc = -1;
>  		} else {
>  promote:
> -			/* We cannot possibly get here in a volatile state. */
> -			assert(box_raft()->volatile_term == box_raft()->term);
> +			if (try_wait)
> +				raft_new_term(box_raft());

2. It starts to bother me, that we use try_wait flag as a kind of a
state of the promote process rather just just a command 'you need to wait'.
But I can't propose anything better right away. Just noticing.

> diff --git a/test/replication/gh-4114-local-space-replication.result b/test/replication/gh-4114-local-space-replication.result
> index 9b63a4b99..676400cef 100644
> --- a/test/replication/gh-4114-local-space-replication.result
> +++ b/test/replication/gh-4114-local-space-replication.result> @@ -77,9 +76,9 @@ box.space.test:insert{3}
>   | - [3]
>   | ...
>  
> -box.info.vclock[0]
> +box.info.vclock[0] == a + 3 or box.info.vclock[0] - a

3. Maybe use an assertion? They really do look easier to read
when you try to understand a test.

>   | ---
> - | - 3
> + | - true
>   | ...
> diff --git a/test/replication/gh-6034-promote-bump-term.result b/test/replication/gh-6034-promote-bump-term.result
> new file mode 100644
> index 000000000..20e352922
> --- /dev/null
> +++ b/test/replication/gh-6034-promote-bump-term.result
> @@ -0,0 +1,37 @@
> +-- test-run result file version 2
> +test_run = require('test_run').new()
> + | ---
> + | ...
> +
> +-- gh-6034: test that every box.ctl.promote() bumps
> +-- the instance's term. Even when elections are disabled. Even for consequent
> +-- promotes on the same instance.
> +election_mode = box.cfg.election_mode
> + | ---
> + | ...
> +box.cfg{election_mode='off'}
> + | ---
> + | ...
> +
> +term = box.info.election.term
> + | ---
> + | ...
> +box.ctl.promote()
> + | ---
> + | ...
> +assert(box.info.election.term == term + 1)
> + | ---
> + | - true
> + | ...
> +box.ctl.promote()
> + | ---
> + | ...
> +assert(box.info.election.term == term + 2)
> + | ---
> + | - true

4. Could you please remind why do we issue a new promote even
if we own the limbo already?

Especially if its ownership is going to get ever more strict
after this series.