Thanks Jamie. This is helpful.

Do we need to start the job again if cluster is restarted due to maintenance or 
upgrades ?

Sent from my iPhone

> On Oct 24, 2019, at 11:58 AM, Jamie <jamied...@aol.co.uk.invalid> wrote:
> 
> Hi Asmath
> When you create a connector the request is forwarded onto the rest api of the 
> worker which is the leader, it uses a Kafka producer to update the topic 
> which stores your connector config. The other workers will continuously 
> monitor this config topic and know if it changes then then they need to 
> perform a rebalance since a new connector has been added. When your workers 
> start they will read the config from this toopic and keep it in memory so 
> they know the current configuration of the connectors. This topic and the 
> other internal topics used for Kafka connect should use compaction so that 
> they aren't removed after the retention time, the most recent version of each 
> connectors config remains.
> For alerting specifically on connector status you could use the REST API to 
> get the status of each connector and it's tasks at regular intervals and 
> alert if the status is failed on any (or define threshold of an acceptable 
> number of failures )
> Hope this helps, 
> Jamie
> 
> 
> -----Original Message-----
> From: KhajaAsmath Mohammed <mdkhajaasm...@gmail.com>
> To: users <users@kafka.apache.org>
> Sent: Thu, Oct 24, 2019 03:52 PM
> Subject: Monitor Kafka connect jobs
> 
> 
> Hi,
> 
> 
> We are using kafka connect in production and I have few questions about it.
> when we submit kafka connect job using rest api . Job gets continously
> running in the background and due to some issues, lets say we restarted
> kafka cluster. do we need to start manually all the jobs again?
> 
> 
> Is there a way to monitor these jobs using tools. I know we can use connect
> UI but in case if we have more than 1000 jobs it would become more complex.
> 
> 
> I am also looking for trigger mechanism to send out email or alert support
> team if the connector was killed due to some reasons.
> 
> 
> Thanks,
> 
> Asmath

Reply via email to