Scenario with ZK ensemble partitioned from Kafka cluster + ISR broker going down

2017-06-15 Thread Kostas Christidis
Assume: 1. A Kafka cluster with 3 brokers: B0, B1, B2 2. A single topic with a single partition 3. default.replication.factor = 3 4. min.insync.replicas = 3 5. acks = all for the producers Let: 1. B0 be the controller of the cluster and the leader of the partition for simplicity 2. ISR = [B0 B2]

What happens to a partition leader whose connection to ZK breaks?

2016-10-21 Thread Kostas Christidis
Does it stop acting as the leader (i.e. stop serving produce and fetch requests) returning the "not a leader for partition" exception? Or does it keep thinking it's the leader? If it's the latter, any connected consumers that wait for new requests on that replica will do so in vain. Since the clus

How much time does it usually take for a rolling upgrade to complete?

2018-07-02 Thread Kostas Christidis
The answer to this depends on several factors: 1. Broker specs, 2. Connectivity between cluster brokers, 3. Number of messages stored in each partition, etc. I'd still like to get a sense how long rolling upgrades take for deployments out there however. My setup: 1. 5-broker cluster with 10GBp

Broker serving bad metadata (Kafka 1.0.0)

2018-09-04 Thread Kostas Christidis
I'm reasonably certain I've bumped into a case where a broker in my cluster is serving bad metadata to clients, making them think that a certain partition is leaderless. This is similar to what's described here (end of the post): https://github.com/Shopify/sarama/issues/665#issuecomment-223128098

Re: Broker serving bad metadata (Kafka 1.0.0)

2018-09-04 Thread Kostas Christidis
Follow-up question: 3. If this is indeed an issue, how does it happen? What makes the broker go into that state? On Tue, Sep 4, 2018 at 10:14 AM Kostas Christidis wrote: > > I'm reasonably certain I've bumped into a case where a broker in my > cluster is serving bad metadata

Relationship fetch.replica.max.bytes and message.max.bytes

2017-03-22 Thread Kostas Christidis
Can fetch.replica.max.bytes be equal to message.max.bytes? 1. The defaults in the official Kafka documentation [1] have the parameter "fetch.replica.max.bytes" set to a higher value than "message.max.bytes". However, nothing in the description of these parameters implies that equality would be wro

Re: Relationship fetch.replica.max.bytes and message.max.bytes

2017-03-23 Thread Kostas Christidis
On Thu, Mar 23, 2017 at 5:04 AM, Ben Stopford wrote: > Hi Kostas - The docs for replica.fetch.max.bytes should be helpful here: > > The number of bytes of messages to attempt to fetch for each partition. > This is not an absolute maximum, if the first message in the first > non-empty partition of

Re: Relationship fetch.replica.max.bytes and message.max.bytes

2017-03-24 Thread Kostas Christidis
max.bytes in your broker config to be " + > "equal or larger than your settings for max.message.bytes, both at > a broker and topic level.") > > Ismael > > On Thu, Mar 23, 2017 at 2:37 PM, Kostas Christidis wrote: > >> On Thu, Mar 23, 2017 at 5:04 A