[ClusterLabs] Antw: [EXT] Re: temporary loss of quorum when member starts to rejoin

[Sherrard Burton]

On 4/7/20 2:50 AM, Ulrich Windl wrote:
>>>> Andrei Borzenkov <arvidjaar at gmail.com> schrieb am 06.04.2020 um 22:10 in
> Nachricht
> <17546_1586203904_5E8B8D00_17546_12_1_73cdd72d-c884-05a4-6c64-2e354912c28f at gmail
> com>:
> 
> [...]
>> I cannot reproduce it, but I also do not use knet. From documentation I
>> have impression that knet has artificial delay before it considers links
>> operational, so may be that is the reason.
> [...]
> 
> NICs may behave quite different: I can remember som early 1Gb model needing almost 5 seconds to negotiate a link with the switch, and the link went up/down at least twice while doing so. Recent NICs seem somewhat faster, but I think something in the 3-second area is still realistic. Maybe software waits a short while until it trusts the NIC status. I don't know...

Ulrich,
it doesn't take quite that long for KNET to "settle", but critically it 
appears to take longer than the qnet negotiation.