diennea/blazingcache

Concurrent fetch&loads can lead to an unstable unresponsive cache

Opened this issue · 0 comments

diegosalvi commented

BUG REPORT

  • Please describe the issue you observed:
    In a production environment I had many cache client stuck in fetch or invalidate calls for the same key. Each client was "competing" for the same key locking the other ones.

  • What did you do?
    To resolve I just needed to stop cache requests on such stuck key. The concurrency problem resolved itself and restarting cache request on such key it didn't happen anymore

  • What did you expect to see?
    Concurrent fetch&invalidate from different clients shouldn't lock themselves.

  • What did you see instead?
    Cache clients were unresponsive on stuck key (waiting till timeout) and on cache server i found many
    broacast pending #17023914 invalidateKey stukkey from java@server1_13152799140601533 started at 2023-01-26 11:01:27.523, remaining [java@server2_13152759600094298, java@server3_13153278964203632