Binbin
> The PR title says "Optimize ..." but it is more than an optimization. Actually a bug fix? Please improve the title. :) I think it is indeed more of...
i try to fix the test in #1288
@PingXie This is a change I wanted to make a while ago, we'll discuss it in a later version.
The motivation is not to exit immediately because it is passive (if it is a primary it may cause the cluster to go down and be unable to write for...
> I've also see issues with stuck IO causing the fsync to stall, which also can cause an outage. this is also a problem, i think we should figure out...
I'd really appreciate if you guys have some input on this one @zuiderkwast @hpatro @PingXie @madolson
> If saving nodes.conf succeeds once and later it fails, then next time we reboot we can end up with an outdated cluster config. I think this is a problem?...
> The "diskless cluster mode," in my view, takes a different approach by delegating configuration management to an external component. However, I don’t think it would work for users who...
> The async write is just a means for the server to be able to time out stuck nodes.conf file updates. The update semantics should still be sync on a...
another thought i want to mention is that, a lot of normal users (not the self-contained one) think Redis/Valkey is a memory databases, when disabling the persistence (RDB and AOF),...