Hi Robert, By colocating the zookeeper servers on the same VM we could have a downtime if the VM happened to go down. Having the servers on different VMs will reduce the risk factor.
For example, if the zookeeper cluster has 5 servers you can tolerate upto 2 servers going down and still have no downtime. Regards, Srinath On Jul 27, 2017 5:55 AM, "Robert Friberg" <robert.frib...@devrex.se> wrote: Hi, I'm preparing a production grade install of Kafka. The message volume will be very low, peak load in messages/second would be measured in the 100's. Message size is on average 0.3kb. The number of topics will be < 100 and the typical number of consumers per topic 2 or 3. Reading the docs and other best practices found it seems like a lot of the recommendations are focused on performance. I'm guessing that some practices are more important than others. In particular I'm wondering if it would be suitable to co-locate the zookeepers on the same VM's as I've heard from multiple sources that this is not recommended. My guess is that this is acceptable given the small workload. Does anyone here disagree or care to elaborate why this would be bad/good? -- Robert Friberg Building OrigoDB at Devrex Labs +46733839080
