Coordinate Sampler Thread across a Checkpoint/Restore

[dsouzai]

• Coordinate the Sampler Thread across a Checkpoint/Restore
• Reset the Start and Elapsed Time in the Restore Hook

Part of #16853

[dsouzai]

@mpirvu could you please review?

[dsouzai]

We will need to update the documentation regarding -XsamplingExpirationTime, but because the CRIU feature hasn't GA'd yet, it should be fairly straightforward. Given that 0.43 has already split, this is something we'll need to do for 0.44.

@mpirvu good for review again. Testing shows everything behaves as expected:

#CHECKPOINT RESTORE: Preparing for checkpoint
#CHECKPOINT RESTORE: Preparing to compile methods for checkpoint
...
#CHECKPOINT RESTORE: Done compiling methods for checkpoint
#CHECKPOINT RESTORE: Preparing to suspend threads for checkpoint
#CHECKPOINT RESTORE: Finished suspending threads for checkpoint
#CHECKPOINT RESTORE: Suspending Sampler Thread for Checkpoint
#CHECKPOINT RESTORE: Ready for checkpoint
#CHECKPOINT RESTORE: Preparing for restore
#CHECKPOINT RESTORE: Start and elapsed time: startTime=1245049424, elapsedTime=  2740
#CHECKPOINT RESTORE: Reset start and elapsed time: startTime=1245074324, elapsedTime=  2740
#CHECKPOINT RESTORE: Resuming Sampler Thread from Checkpoint
#CHECKPOINT RESTORE: Resetting Sampling Thread Lifetime State
#CHECKPOINT RESTORE: Ready for restore

[mpirvu]

LGTM. I have a couple of comments inline.
Also, there is the potential for the checkpoint to wait for an entire sampler period. If the checkpointing thread sends the interrupt when the sampler is not sleeping, the sampler will continue to run in its loop and then start sleeping. It will pick up the checkpoint intention after the sleep is over. Given that the sleep time is ~10 ms, this should not be a big problem.

[mpirvu]

jenkins test sanity all jdk17

[dsouzai]

I believe AIX failure is #8625

[dsouzai]

jenkins test sanity aix jdk17