Closed pshipton closed 5 years ago
@Mesbah-Alam
@Mesbah-Alam any outlook for getting this fixed? It's causing all the extended system testing to fail across all platforms.
I am looking at the test right now.
This test has 4 steps:
* ~~~~~~~~~~~~~~~~~~~~~~
* 1) Start Jvm1 with options: "-XX:SharedCacheHardLimit=20m, -Xscmxsoftmx=1m". It should
* create the shared classes cache and run a mini-mix load test and fill up the shared
* classes cache.
* 2) Once Jvm1 finishes its work, check cache stats to ensure cache is soft full. Fail if it isn't.
* 3) Start Jvm2 that does -Xshareclasses:adjustsoftmx=2m. Check cache stats to ensure
* cache's softmx limit has been increased to 2m.
* 4) Start Jvm3 that runs the second mini-mix load. Make sure code is being written to
* the increased space of the shared classes cache.
The test fails in step 2; i.e. it fills up to 99% when it's expected to fill up 100%. We need to retune the test.
example failure: https://ci.eclipse.org/openj9/job/Test_openjdk13_j9_extended.system_x86-64_linux_Nightly/17 for all failures see https://ci.eclipse.org/openj9/job/Pipeline-Build-Test-All/620/
SC_Softmx_Increase_0 is failing on what appears to be all platforms and java versions after https://github.com/eclipse/openj9/pull/6536 is merged. This PR changes the format of the shared cache and consumes more space. I expect the test needs to be re-tuned.