not. Without this guarantee, it would have to be called ACI.
>
> Eugen
>
>
> 差出人: Alexandre Dupriez
> 送信日時: 2020年3月8日 0:10
> 宛先: users@kafka.apache.org
> 件名: Re: synchronously flushing messages to disk
>
> Hi Eugen,
>
> The first lin
all, this is what all ACID databases rely on,
replicated or not. Without this guarantee, it would have to be called ACI.
Eugen
差出人: Alexandre Dupriez
送信日�r: 2020年3月8日 0:10
宛先: users@kafka.apache.org
件名: Re: synchronously flushing messages to disk
Hi Eugen,
The
Hi Eugen,
The first line of config log.flush.interval.messages=1 will make Kafka
force an fsync(2) for every produce requests.
The second line of config is not sufficient for periodic flush, you
also need to update log.flush.scheduler.interval.ms which is Long.Max
by default (in which case period-
I was under the impression that these settings
log.flush.interval.messages=1
log.flush.interval.ms=0
guarantee a synchronous fsync for every message, i.e.when the producer receives
an ack for a message, it is guaranteed to have been persisted to as many disks
as min.insync.replicas requires.
As
