[Tarantool-patches] [PATCH 2/2] raft: fix crash on candidate cfg during WAL write

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Tue Nov 10 01:42:12 MSK 2020


Thanks for the review!

>> diff --git a/test/replication/gh-5506-election-on-off.result b/test/replication/gh-5506-election-on-off.result
>> index 1a718396f..4fbc31986 100644
>> --- a/test/replication/gh-5506-election-on-off.result
>> +++ b/test/replication/gh-5506-election-on-off.result
>> @@ -47,6 +47,91 @@ box.error.injection.set("ERRINJ_WAL_DELAY", false)
>>    | - ok
>>    | ...
>>   +box.cfg{                                                                        \
>> +    election_mode = old_election_mode,                                          \
>> +}
>> + | ---
>> + | ...
>> +
>> +--
>> +-- Another crash could happen when election mode was configured to be
>> +-- 'candidate' with a known leader, but there was a not finished WAL write.
>> +-- The node tried to start waiting for the leader death, even though with an
>> +-- active WAL write it should wait for its end first.
>> +--
>> +box.schema.user.grant('guest', 'super')
> 
> 
> box.schema.user.grant('guest', 'replication') is enough.

I know. But I prefer using super, because it works always. I don't
want to think of the weakest possible rights which would fit. So
I usually add 'super' and forget about it.

>> + | ---
>> + | ...
>> +test_run:cmd('create server replica with rpl_master=default,\
>> +              script="replication/replica.lua"')
>> + | ---
>> + | - true
>> + | ...
>> +test_run:cmd('start server replica with wait=True, wait_load=True')
>> + | ---
>> + | - true
>> + | ...
>> +
>> +test_run:switch('replica')
>> + | ---
>> + | - true
>> + | ...
>> +box.cfg{election_mode = 'voter'}
>> + | ---
>> + | ...
>> +box.error.injection.set("ERRINJ_WAL_DELAY_COUNTDOWN", 0)
>> + | ---
>> + | - ok
>> + | ...
>> +
>> +test_run:switch('default')
>> + | ---
>> + | - true
>> + | ...
>> +box.cfg{election_mode = 'candidate'}
>> + | ---
>> + | ...
>> +test_run:wait_cond(function()                                                   \
>> +    return box.info.election.leader ~= 0                                        \
>> +end)
> 
> 
> Wouldn't it be simpler to wait for `box.info.election.state == 'leader'`?

I don't mind.

====================
diff --git a/test/replication/gh-5506-election-on-off.result b/test/replication/gh-5506-election-on-off.result
index 4fbc31986..b8abd7ecd 100644
--- a/test/replication/gh-5506-election-on-off.result
+++ b/test/replication/gh-5506-election-on-off.result
@@ -92,7 +92,7 @@ box.cfg{election_mode = 'candidate'}
  | ---
  | ...
 test_run:wait_cond(function()                                                   \
-    return box.info.election.leader ~= 0                                        \
+    return box.info.election.state == 'leader'                                  \
 end)
  | ---
  | - true
diff --git a/test/replication/gh-5506-election-on-off.test.lua b/test/replication/gh-5506-election-on-off.test.lua
index bb89477d1..476b00ec0 100644
--- a/test/replication/gh-5506-election-on-off.test.lua
+++ b/test/replication/gh-5506-election-on-off.test.lua
@@ -47,7 +47,7 @@ box.error.injection.set("ERRINJ_WAL_DELAY_COUNTDOWN", 0)
 test_run:switch('default')
 box.cfg{election_mode = 'candidate'}
 test_run:wait_cond(function()                                                   \
-    return box.info.election.leader ~= 0                                        \
+    return box.info.election.state == 'leader'                                  \
 end)
 
 test_run:switch('replica')


More information about the Tarantool-patches mailing list