valitydev/riak_test
14
0
Fork 0
mirror of https://github.com/valitydev/riak_test.git synced 2024-11-06 08:35:22 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fix false positive test fails caused by AAE race

Browse Source 
It is occasionally possible to see a YZ AAE repair in normal operation
if we happen to be running the exchange FSM while writes are occurring,
and we compare the KV tree (which already has the new data inserted)
with the YZ tree (which doesn't have the data inserted yet).

This ends up triggering an extra call to get_map, which could cause some
earlier parts of this test to fail when it expected 60 calls but saw 61.

To fix this, we can simply disable AAE at the start of the test, and
then enable it later when we get to the part of the test that requires
AAE.
This commit is contained in:
Nick Marino 2016-03-30 14:35:46 -04:00
parent 951473a3d7
commit 85eaf86f81
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tests/yz_extractors.erl
View File

@ -145,7 +145,9 @@
%% allow AAE to build trees and exchange rapidly
{anti_entropy_build_limit, {100, 1000}},
{anti_entropy_concurrency, 8},
{anti_entropy_tick, 1000}
{anti_entropy_tick, 1000},
%% but start with AAE turned off so as not to interfere with earlier parts of the test
{anti_entropy, {off, []}}
]},
{yokozuna,
[
@ -318,6 +320,8 @@ test_extractor_with_aae_expire(Cluster, Index, Bucket, Packet) ->
yokozuna_rt:search_expect({Host, Port}, Index, <<"host">>,
<<"www*">>, 1),
rpc:multicall(Cluster, riak_kv_entropy_manager, enable, []),
yokozuna_rt:expire_trees(Cluster),
yokozuna_rt:wait_for_full_exchange_round(Cluster, erlang:now()),