You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I work on branch develop-3.0 but this issue seems not related to specifically that branch.
Looking into the QuickCheck tests that actually fail, but are masked for failure by ?TRAPEXIT:
Possibly this is just an artefact on how the property is written
When calling riak_core_bg_manager:start() two ets tables are generated before actually starting:
These ets tables are supposed to survive a crash of riak_core_bg_manager, such that after restart, these ets tables contain the process that hold locks.
The error I see in the QuickCheck tests is that aparently when a process monitored by the manager dies normally, it may not be able to find it in the ets table and causes a crash of the manager.
=ERROR REPORT==== 15-Mar-2019::13:17:52 ===
** Generic server riak_core_bg_manager terminating
** Last message in was {'DOWN',#Ref<0.2179874273.4102029316.175607>,process,
<0.2248.0>,normal}
** When Server state == {state,background_mgr_info_table,
background_mgr_entry_table,true,false}
** Reason for termination ==
** {badarg,
[{ets,match_object,[background_mgr_entry_table,{{given,'_'},'_'}],[]},
{riak_core_bg_manager,release_resource,2,
[{file,
"/Users/thomas/Quviq/.../riak_core/_build/eqc/lib/riak_core/src/riak_core_bg_manager.erl"},
{line,707}]},
{riak_core_bg_manager,handle_info,2,
[{file,
"/Users/thomas/Quviq/.../riak_core/_build/eqc/lib/riak_core/src/riak_core_bg_manager.erl"},
{line,505}]},
{gen_server,try_dispatch,4,[{file,"gen_server.erl"},{line,616}]},
{gen_server,handle_msg,6,[{file,"gen_server.erl"},{line,686}]},
{proc_lib,init_p_do_apply,3,[{file,"proc_lib.erl"},{line,247}]}]}
I am going to dive deeper into this soon, but if people have seen this before or have suggestions, please let me know.
The text was updated successfully, but these errors were encountered:
I work on branch
develop-3.0
but this issue seems not related to specifically that branch.Looking into the QuickCheck tests that actually fail, but are masked for failure by ?TRAPEXIT:
Possibly this is just an artefact on how the property is written
When calling riak_core_bg_manager:start() two ets tables are generated before actually starting:
These ets tables are supposed to survive a crash of
riak_core_bg_manager
, such that after restart, these ets tables contain the process that hold locks.The error I see in the QuickCheck tests is that aparently when a process monitored by the manager dies normally, it may not be able to find it in the ets table and causes a crash of the manager.
I am going to dive deeper into this soon, but if people have seen this before or have suggestions, please let me know.
The text was updated successfully, but these errors were encountered: