Hi Alan,
Oh, we just have 2G swap buffer on our machine.
Yes, we do have observed some other intermittent failures due to
memory/network/GUI reasons, but this case is the most frequently
observed one.
I just try to improve the test stability since it seems impossible to
always keep others away from the machine when doing the tests.
Thanks.
Best regards,
Jie
On 2019/5/18 下午4:25, Alan Bateman wrote:
On 18/05/2019 02:17, Jie Fu wrote:
Hi Alan,
On 2019/5/17 下午11:00, Alan Bateman wrote:
The issue isn't clear. The test only runs when on 64-bit systems
with >= 10GB memory. Is the issue related to the -concurrency
setting that you specify to jtreg?
No. Even we set the concurrency=1, the issue may still occur.
os.maxMemory >= 10g just check whether the capacity of physical
memory is >= 10g, not the available memory to be used.
Our test machine has only 16GB physical memory and is shared by many
people in our group.
The test seems to fail occasionally especially when the test machine
is busy.
I'm trying to find a better solution to this problem.
I'm guessing this must be an embedded system with no swap configured.
If so then I don't think this is solvable without reserving the
machine when you need to run tests. This test uses 6-8GB of memory and
there are several other tests that consume lots of memory too. If
several people are trying to run tests or do builds at the same time
then I assume you'll have issues in many areas. I see your patch to
have the test skip when there isn't 6GB of "free memory" but I assume
you'll still have intermittent failures when someone else tries to run
resource hungry tests at around the same time.
-Alan
