Make FakerRepeatabilityIntegrationTest running in same thread

[snuyanzin]

No description provided.

[kingthorin]

Is this instead of the other PR or in addition?

[snuyanzin]

addition to fix repeatability

[bodiam]

I don't see a reason for this. This code worked fine before d9c2cc262d29deb3552c9f49ab71edf0b3f6dc30.

This seems to hide that we still have a concurrency issue.

[snuyanzin]

how do you that exactly d9c2cc2 is the reason?
after that there were about 10 more commits to main branch and no such issue

[bodiam]

how do you that exactly d9c2cc2 is the reason? after that there were about 10 more commits to main branch and no such issue

Because I checked out all these commits:

And all of them work consistently. Except for d9c2cc262d29deb3552c9f49ab71edf0b3f6dc30. If I added a repeatedtest on the method there, things start failing.

d9c2cc262d29deb3552c9f49ab71edf0b3f6dc30: fails
b840846fcfbf88112241b398faced21903435232: works
1e7544bd78a11563bed22c25af91f550a90d52ed: works
9a32f4b0933c74f3e1298a6a36341a6301b2e09f: works

update:

6ee07320144dbba7ddf06e1fbd62db4ce6bd8a4c: works
64e52940983202fbb0387044491ef8178ac46bd6: works
61eb73fa4a7bc5ae718b582b69d4d998c2be1ea6: works
a3f9995b21ee928bd17c828720a16817aad5b9fd: works
88c0e389a27775f8aa90df6c7a6e79051af7d851: works
b2cb6a803a46e7a82535cfb3add607098f62439d: FAIL

In the main branch, things started failing since the Azure Spring Apps are there??

[bodiam]

So, that's interesting: if we enabled the springApps() method in Azure, things started failing.

[snuyanzin]

i wonder if it's because of springApps() or just because of any 1 more method which means that if we still have enabled springApps() however remove any another method will it work?

[bodiam]

i wonder if it's because of springApps() or just because of any 1 more method which means that if we still have enabled springApps() however remove any another method will it work?

I don't think it's because of springApps per se, it's not doing much, but it seems that just calling the randHex method is triggering the issue.

I've done a few random tests, and the funny things is:

    public String springApps() {
        String hex = faker.random().hex(4, false); // works
        return "sa-" + hex;
    }

This consistently works.

    public String springApps() {
        String hex = faker.random().hex(8, false); // works
        return "sa-" + hex;
    }

but this fails:

    public String springApps() {
        String hex = faker.random().hex(9, false); // 9 fails. Anything above fails. 8 = 1 byte, 9 = 2 bytes?
        return "sa-" + hex;
    }

But I don't understand it. All the other methods are using the same code, but with 16, yet they don't fail, but this does. I'm probably chasing ghosts here?

[snuyanzin]

After some tests it seems springApps is only a trigger and not the rootcause. The reason is CopyOnWriteMap introduced in #770
Where in case of high throughput some values could be removed and recalculated once again (normal behavior of copy on write)

well I still suggest to make repeatabale tests in same thread.

Before that fix we anyway didn't support multithreading for FakeValuesService as mentioned at #759

[bodiam]

well I still suggest to make repeatabale tests in same thread.

Okay, that's fine!