Hi Chris, Good KIP – I think it will be very helpful in alerting and automating the resolution of common Connect problems.
I have a couple of questions / suggestions: 1. What are you planning on documenting as guidance for using this new endpoint? My guess would be that if Connect doesn’t return a status of 200 after some period I would either page someone, or restart the process? But I’m missing the nuance of distinguishing between readiness and liveness, is this for maintaining overall availability when rolling out updates to several Connect processes? 2. Would you consider providing a way to discover details about exactly which condition (or conditions) is/are failing? Perhaps by providing a richer JSON response? Something like this would help us adopt the health check, as we could start by paging someone for all failures, then over time (as we gained confidence) transition more of the conditions over to being handled by automation. Regards, - Adrian From: Chris Egerton <chr...@aiven.io.INVALID> Date: Monday, 10 June 2024 at 15:26 To: dev@kafka.apache.org <dev@kafka.apache.org> Subject: [EXTERNAL] Re: [DISCUSS] KIP-1017: A health check endpoint for Kafka Connect Hi all, Thanks for the positive feedback! I've made one small addition to the KIP since there's been a change to our REST timeout error messages that's worth incorporating here. Quoting the added section directly: > Note that the HTTP status codes and "status" fields in the JSON response will match the exact examples above. However, the "message" field may be augmented to include, among other things, more information about the operation(s) the worker could be blocked on (such as was added in REST timeout error messages in KAFKA-15563 [1]). Assuming this still looks okay to everyone, I'll kick off a vote thread sometime this week or the next. [1] - https://issues.apache.org/jira/browse/KAFKA-15563 Cheers, Chris On Fri, Jun 7, 2024 at 12:01 PM Andrew Schofield <andrew_schofi...@live.com> wrote: > Hi Chris, > This KIP looks good to me. I particularly like the explanation of how the > result will specifically > check the worker health in ways that have previously caused trouble. > > Thanks, > Andrew > > > On 7 Jun 2024, at 16:18, Mickael Maison <mickael.mai...@gmail.com> > wrote: > > > > Hi Chris, > > > > Happy Friday! The KIP looks good to me. +1 > > > > Thanks, > > Mickael > > > > On Fri, Jan 26, 2024 at 8:41 PM Chris Egerton <chr...@aiven.io.invalid> > wrote: > >> > >> Hi all, > >> > >> Happy Friday! I'd like to kick off discussion for KIP-1017, which (as > the > >> title suggests) proposes adding a health check endpoint for Kafka > Connect: > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1017%3A+Health+check+endpoint+for+Kafka+Connect > >> > >> This is one of the longest-standing issues with Kafka Connect and I'm > >> hoping we can finally put it in the ground soon. Looking forward to > hearing > >> people's thoughts! > >> > >> Cheers, > >> > >> Chris > > Unless otherwise stated above: IBM United Kingdom Limited Registered in England and Wales with number 741598 Registered office: PO Box 41, North Harbour, Portsmouth, Hants. PO6 3AU
