[Tarantool-patches] [PATCH 1/3 v2] replication: test clear_synchro_queue() function

[Serge Petrenko]

12.11.2020 12:10, sergeyb at tarantool.org пишет:
> From: Sergey Bronnikov <sergeyb at tarantool.org>
>
> Part of #5055
> Part of #4849

Hi! Thanks for the patch!

Please see 2 comments below.

> ---
>   test/replication/qsync_basic.result   | 136 ++++++++++++++++++++++++++
>   test/replication/qsync_basic.test.lua |  50 ++++++++++
>   2 files changed, 186 insertions(+)
>
> diff --git a/test/replication/qsync_basic.result b/test/replication/qsync_basic.result
> index bd3c3cce1..e848e305a 100644
> --- a/test/replication/qsync_basic.result
> +++ b/test/replication/qsync_basic.result
> @@ -32,6 +32,13 @@ s2.is_sync
>    | - false
>    | ...
>   
> +--
> +-- gh-4849: clear synchro queue with unconfigured box
> +--

What do you mean by `unconfigured box`?

box.cfg is called when the default instance (master.lua) is started, before
the first test line.

If you really  want to test `box.ctl.clear_synchro_queue` before box.cfg,
you need to do it in `app-tap/cfg.test.lua`. It has all the tests for 
`box.ctl` options.

> +box.ctl.clear_synchro_queue()
> + | ---
> + | ...
> +
>   -- Net.box takes sync into account.
>   box.schema.user.grant('guest', 'super')
>    | ---
> @@ -637,6 +644,135 @@ box.space.sync:count()
>    | - 0
>    | ...
>   
> +--
> +-- gh-4849: clear synchro queue on a master
> +--
> +test_run:switch('default')
> + | ---
> + | - true
> + | ...
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout = 1000}
> + | ---
> + | ...
> +ok, err = nil
> + | ---
> + | ...
> +f = fiber.create(function()							\
> +    ok, err = pcall(box.space.sync.insert, box.space.sync, {10})		\
> +end)
> + | ---
> + | ...
> +f:status()
> + | ---
> + | - suspended
> + | ...
> +test_run:switch('replica')
> + | ---
> + | - true
> + | ...
> +test_run:wait_cond(function() return box.space.sync:get{10} ~= nil end)
> + | ---
> + | - true
> + | ...
> +test_run:switch('default')
> + | ---
> + | - true
> + | ...
> +box.cfg{replication_synchro_timeout = 0.1}
> + | ---
> + | ...
> +box.ctl.clear_synchro_queue()
> + | ---
> + | ...
> +test_run:switch('replica')
> + | ---
> + | - true
> + | ...
> +test_run:wait_cond(function() return box.space.sync:get{10} == nil end)
> + | ---
> + | - true
> + | ...
> +test_run:switch('default')
> + | ---
> + | - true
> + | ...
> +test_run:wait_cond(function() return f:status() == 'dead' end)
> + | ---
> + | - true
> + | ...
> +ok, err
> + | ---
> + | - false
> + | - Quorum collection for a synchronous transaction is timed out
> + | ...
> +test_run:wait_cond(function() return box.space.sync:get{10} == nil end)
> + | ---
> + | - true
> + | ...
> +
> +--
> +-- gh-4849: clear synchro queue on a replica, make sure no crashes
> +--
> +test_run:switch('default')
> + | ---
> + | - true
> + | ...
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout = 1000}
> + | ---
> + | ...
> +ok, err = nil
> + | ---
> + | ...
> +f = fiber.create(function()                                                     \
> +    ok, err = pcall(box.space.sync.insert, box.space.sync, {9})                 \
> +end)
> + | ---
> + | ...
> +f.status()
> + | ---
> + | - running
> + | ...


typo: should be `f:status()`


> +test_run:wait_cond(function() return box.space.sync:get{9} ~= nil end)
> + | ---
> + | - true
> + | ...
> +test_run:switch('replica')
> + | ---
> + | - true
> + | ...
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout=0.01}
> + | ---
> + | ...
> +box.ctl.clear_synchro_queue()
> + | ---
> + | ...
> +test_run:wait_cond(function() return box.space.sync:get{9} == nil end)
> + | ---
> + | - true
> + | ...
> +test_run:switch('default')
> + | ---
> + | - true
> + | ...
> +box.cfg{replication_synchro_timeout=0.01}
> + | ---
> + | ...
> +test_run:wait_cond(function() return f:status() == 'dead' end)
> + | ---
> + | - true
> + | ...
> +ok, err
> + | ---
> + | - false
> + | - Quorum collection for a synchronous transaction is timed out
> + | ...
> +test_run:wait_cond(function() return box.space.sync:get{9} == nil end)
> + | ---
> + | - true
> + | ...
> +
> +-- Note: cluster may be in a broken state here due to nature of previous test.
> +
>   -- Cleanup.
>   test_run:cmd('switch default')
>    | ---
> diff --git a/test/replication/qsync_basic.test.lua b/test/replication/qsync_basic.test.lua
> index 94235547d..523dfa779 100644
> --- a/test/replication/qsync_basic.test.lua
> +++ b/test/replication/qsync_basic.test.lua
> @@ -13,6 +13,11 @@ s1:select{}
>   s2 = box.schema.create_space('test2')
>   s2.is_sync
>   
> +--
> +-- gh-4849: clear synchro queue with unconfigured box
> +--
> +box.ctl.clear_synchro_queue()
> +
>   -- Net.box takes sync into account.
>   box.schema.user.grant('guest', 'super')
>   netbox = require('net.box')
> @@ -248,6 +253,51 @@ for i = 1, 100 do box.space.sync:delete{i} end
>   test_run:cmd('switch replica')
>   box.space.sync:count()
>   
> +--
> +-- gh-4849: clear synchro queue on a master
> +--
> +test_run:switch('default')
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout = 1000}
> +ok, err = nil
> +f = fiber.create(function()							\
> +    ok, err = pcall(box.space.sync.insert, box.space.sync, {10})		\
> +end)
> +f:status()
> +test_run:switch('replica')
> +test_run:wait_cond(function() return box.space.sync:get{10} ~= nil end)
> +test_run:switch('default')
> +box.cfg{replication_synchro_timeout = 0.1}
> +box.ctl.clear_synchro_queue()
> +test_run:switch('replica')
> +test_run:wait_cond(function() return box.space.sync:get{10} == nil end)
> +test_run:switch('default')
> +test_run:wait_cond(function() return f:status() == 'dead' end)
> +ok, err
> +test_run:wait_cond(function() return box.space.sync:get{10} == nil end)
> +
> +--
> +-- gh-4849: clear synchro queue on a replica, make sure no crashes
> +--
> +test_run:switch('default')
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout = 1000}
> +ok, err = nil
> +f = fiber.create(function()                                                     \
> +    ok, err = pcall(box.space.sync.insert, box.space.sync, {9})                 \
> +end)
> +f.status()
> +test_run:wait_cond(function() return box.space.sync:get{9} ~= nil end)
> +test_run:switch('replica')
> +box.cfg{replication_synchro_quorum = 3, replication_synchro_timeout=0.01}
> +box.ctl.clear_synchro_queue()
> +test_run:wait_cond(function() return box.space.sync:get{9} == nil end)
> +test_run:switch('default')
> +box.cfg{replication_synchro_timeout=0.01}
> +test_run:wait_cond(function() return f:status() == 'dead' end)
> +ok, err
> +test_run:wait_cond(function() return box.space.sync:get{9} == nil end)
> +
> +-- Note: cluster may be in a broken state here due to nature of previous test.
> +
>   -- Cleanup.
>   test_run:cmd('switch default')
>   

-- 
Serge Petrenko