WIP

[Al2Klimov]

refs #10082

TODO

• Checkable: Don't recalculate next_check for remotely generated cr #10011

[Al2Klimov]

• @yhabteab Are you sure all SetNextCheck remaining after Checkable: Don't recalculate next_check for remotely generated cr #10011 (which don't already call Checkable::OnNextCheckUpdated) shall be written to DB? Especially in Checkable#Start() or those affecting parent/child..

[yhabteab]

No, I'm not! I haven't even checked all of them in detail, but aren't they just like the other superfluous ones from Checkable::ExecuteCheck() and used to enforce the scheduler to re-index its idle checkables queue?

[Al2Klimov]

I don't think so.

icinga2/lib/icinga/checkable-check.cpp

Lines 392 to 404 in bca1a84

	if (recovery) {
	for (auto& child : children) {
	if (child->GetProblem() && child->GetEnableActiveChecks()) {
	auto nextCheck (now + Utility::Random() % 60);
	ObjectLock oLock (child);
	if (nextCheck < child->GetNextCheck()) {
	child->SetNextCheck(nextCheck);
	}
	}
	}
	}

icinga2/lib/icinga/checkable-check.cpp

Lines 406 to 420 in bca1a84

	if (stateChange) {
	/* reschedule direct parents */
	for (const Checkable::Ptr& parent : GetParents()) {
	if (parent.get() == this)
	continue;
	if (!parent->GetEnableActiveChecks())
	continue;
	if (parent->GetNextCheck() >= now + parent->GetRetryInterval()) {
	ObjectLock olock(parent);
	parent->SetNextCheck(now);
	}
	}
	}

Especially these two re-schedule the next check of other checkables as something happened with the current one. The other HA node may be responsible for them, so of course this matters for the cluster. But IMAO this doesn't matter for our backends. I think we can stop our anti-SetNextCheck witch-hunt here.