RFR: 8284900: Check InitialHeapSize and container memory limits before startup [v8]

[Ioi Lam]

On Wed, 20 Apr 2022 16:10:12 GMT, xpbob <duke at openjdk.java.net> wrote:

>> Check InitialHeapSize and container memory limits before startup
>> 
>> 
>> Operating System Metrics:
>>     Provider: cgroupv1
>>     Effective CPU Count: 8
>>     CPU Period: 100000us
>>     CPU Quota: -1
>>     CPU Shares: -1
>>     List of Processors, 8 total: 
>>     0 1 2 3 4 5 6 7 
>>     List of Effective Processors: N/A
>>     List of Memory Nodes, 1 total: 
>>     0 
>>     List of Available Memory Nodes: N/A
>>     Memory Limit: 50.00M
>>     Memory Soft Limit: Unlimited
>>     Memory & Swap Limit: 60.00M
>>     Maximum Processes Limit: 4194305
>> 
>> command:-Xms60m -XshowSettings:system -version 
>> 
>> 
>> 
>> 
>> When the memory limit is exceeded, the startup fail with message
>> 
>> 
>> 
>> Error occurred during initialization of VM
>> Initial heap size set to a larger value than the container memory & swap limit
>
> xpbob has updated the pull request incrementally with one additional commit since the last revision:
> 
>   add test

Changes requested by iklam (Reviewer).

src/hotspot/share/gc/shared/gcArguments.cpp line 149:

> 147:     if ((limit > 0) && (InitialHeapSize >= (julong) limit)) {
> 148:       vm_exit_during_initialization("Initial heap size set to a larger value than the container memory & swap limit");
> 149:     }

Since you are testing `InitialHeapSize >= (julong) limit)`, the error message should be:

> Initial heap size should be smaller than the container memory & swap limit

The JVM will use memory outside of the heap (for example malloc, or code cache), so even if you set the `InitialHeapSize` to below the memory limit, the JVM might still get killed by OOM unpredictably.

This PR catches obvious problems when `-Xms` is set way too large. However, I think we need to consider the larger issue of OOM killer when using Java in containers. I filed [JDK-8285277](https://bugs.openjdk.java.net/browse/JDK-8285277)

I've started a discussion on [hotspot-dev](https://mail.openjdk.java.net/pipermail/hotspot-dev/2022-April/059463.html) about what approach we should use to handle container memory limits. Let's hold on with this PR until we have a consensus.

-------------

PR: https://git.openjdk.java.net/jdk/pull/8256