Add option to skip resolver nameserver request

[sharon-wang]

Java programs run on z/OS are hanging when the resolver cannot contact any of the nameservers. The programs hang until the resolver times out, which is ~60 seconds. By skipping the nameserver request, this situation can be avoided.

Fixes: #4540
Documentation Issue: eclipse-openj9/openj9-docs#226

Signed-off-by: Sharon Wang sharon-wang-cpsc@outlook.com

[pshipton]

You don't need to reproduce the reported delay in order to add the option.

[pshipton]

Please create an issue at https://github.com/eclipse/openj9-docs in order to add the new options to the user guide.

[pshipton]

@llxia can you please review the test changes.

[sharon-wang]

@pshipton

Please create an issue at https://github.com/eclipse/openj9-docs in order to add the new options to the user guide.

The new options will apply to all JDK versions and all platforms?

[pshipton]

The new options will apply to all JDK versions and all platforms?

Correct. Unless you are coding something to restrict either of these (or adding code to a place which is already restricted), OpenJ9 changes apply to all versions and platforms.

[pshipton]

When you have the -verbose:init output, please show the section containing the new message here.

[llxia]

Could we put the test in https://github.com/eclipse/openj9/blob/master/test/functional/cmdLineTests/cmdLineTest_J9tests/j9tests.xml? In this way, we do not need to create separate build.xml and playlist.xml. And the test will run as part of cmdLineTest_J9test_extended https://github.com/eclipse/openj9/blob/b9f8b69153cc7a3c2e183c0b7ae13d819c5441f1/test/functional/cmdLineTests/cmdLineTest_J9tests/playlist.xml#L67.

[sharon-wang]

Could we put the test in https://github.com/eclipse/openj9/blob/master/test/functional/cmdLineTests/cmdLineTest_J9tests/j9tests.xml? In this way, we do not need to create separate build.xml and playlist.xml. And the test will run as part of cmdLineTest_J9test_extended

openj9/test/functional/cmdLineTests/cmdLineTest_J9tests/playlist.xml

Line 67 in b9f8b69
cmdLineTest_J9test_extended
.

Sounds good, I'll move it there

[sharon-wang]

@pshipton Here's the output of java -XX:-ReadIPInfoForRAS -verbose:init -version

Should the prefixed newline be removed and the message be indented so that it lines up with "completed with rc=0 in 34011 nano sec."?

[pshipton]

Should the prefixed newline be removed and the message be indented so that it lines up with "completed with rc=0 in 34011 nano sec."?

Agreed. That would look better.

[sharon-wang]

@pshipton Here's the updated output. Should the first letter be lowercased?

[pshipton]

Should the first letter be lowercased?

sure, that would be consistent with the rest

[pshipton]

After that, if you think this is ready to go, please remove the WIP from the title.

[pshipton]

jenkins test extended plinux jdk8

[pshipton]

The -XX:+ReadIPInfoForRAS -XX:-ReadIPInfoForRAS test is failing because the failure condition string is a subset of the success condition string.

>> Success condition was found: [Output match: not using network query to determine host name and IP address for RAS.]
>> Failure condition was found: [Output match: using network query to determine host name and IP address for RAS.]

[pshipton]

I think the failure condition could just be removed. The test will fail if the success condition is not found.

[sharon-wang]

The -XX:+ReadIPInfoForRAS -XX:-ReadIPInfoForRAS test is failing because the failure condition string is a subset of the success condition string.

I think the failure condition could just be removed. The test will fail if the success condition is not found.

Ah yes I missed that caveat when I lowercased the first letter of each message. I'll remove the failure condition - or, I could change the messages to "enabled/disabled network query to determine host name and IP address for RAS."?

[pshipton]

I could change the messages to "enabled/disabled

Sure. I don't think it makes a big difference either way, as long as the messages are understandable.

[sharon-wang]

@pshipton I ended up changing the messages to include enable/disable. Just finished running the tests to confirm functionality and they passed.

[llxia]

Test changes look good to me.

[sharon-wang]

type="fail" is not supported. "failure" was correct.

That's my bad, I missed that detail. I've changed it back to "failure".

[pshipton]

jenkins test extended plinux jdk11