search

mesos / kafka

Apache Kafka on Apache Mesos
Apache License 2.0
414 stars 140 forks source link

Upgrade instructions from 0.9.1.4 to 0.9.2.0 #132

Closed mindscratch closed 9 years ago

mindscratch commented 9 years ago

I'm currently running 0.9.1.4 and was wondering what steps I'd have to take to upgrade to 0.9.2.0.

It looks like I'd put the new JAR on a server, but do I just need to restart brokers or do I need I remove and add them again?

mindscratch commented 9 years ago

Looking closer, it looks I can just restart the scheduler and that's it. Am I missing anything?

dmitrypekar commented 9 years ago

Hi @mindscratch

You are right. Currently by design you can just stop old scheduler and start a new one. You can also restart each broker for changes in Executor to apply.

Please note: if you are going to do this on prod, note somewhere the current location of brokers to be able to restart them on the same set of slaves. For now the scheduler does support slave stickiness function, but your version could potentially not.

Also please close this ticket.

mindscratch commented 9 years ago

@dmitrypekar I'm currently running 0.9.1.4. Does that version support slave stickiness? If not, how would/could I get the brokers to start on the same hosts as before? Also, to do the restart, would I just restart a single broker, wait a minute, restart the next, etc until all have been restarted? Each of my topics has 3 or more replicas, so I believe I won't lose data.

dmitrypekar commented 9 years ago

@mindscratch yes, stickiness is supported starting from 0.9.1.1. Yes, imho, it is more safe to restart brokers one-by-one. I suppose, that log.dirs is configured to be outside Mesos sandbox? So restarting would use the same log.dirs?

mindscratch commented 9 years ago

yes log.dirs is configured outside the sandbox.

On Tue, Nov 10, 2015 at 8:06 AM, dmitrypekar notifications@github.com wrote:

@mindscratch https://github.com/mindscratch yes, stickiness is supported starting from 0.9.1.1. Yes, imho, it is more safe to restart brokers one-by-one. I suppose, that log.dirs is configured to be outside Mesos sandbox? So restarting would use the same log.dirs?

— Reply to this email directly or view it on GitHub https://github.com/mesos/kafka/issues/132#issuecomment-155414582.

https://github.com/mindscratch https://www.google.com/+CraigWickesser https://twitter.com/mind_scratch https://twitter.com/craig_links

dmitrypekar commented 9 years ago

In that case the above upgrade scenario should work fine. Please report about the results and close the ticket.

dmitrypekar commented 9 years ago

Closing the ticket as outdated.