hanishakoneru commented on pull request #1494: URL: https://github.com/apache/ozone/pull/1494#issuecomment-859696879
> To try this scenario, you can update ozone-ha docker-compose.yaml to use below change This scenario will not work because OM1 was started with all 3 peer information. So it formed a ring assuming there are 3 peers in the system. But since only om1 is active, the Leader Election will never finish. To test bootstrap, we would have to start non-HA or HA regularly and then update the confs to include a new OM and start the new OM in bootstrap mode. > To convert non-HA to HA cluster? (Because for this on a single node we need to enable ratis and restart? And also serviceID is not configured, we use omServiceIdDefault (If we don't specify we use this). And also we need to download snapshot outside ratis, because there are some transactions that happened outside of ratis? As even though we enable ratis, new transactions will be in ratis log, and new OM is bootstrapped it only gets these new logs (If not purged, but it does not get DB, so that is why we need to download snapshot outside ratis) Yes. I will open a new Jira to address this scenario. We could transfer a checkpoint to a bootstrapping OM before the setConf request (but only when there are non-ratis transactions in RocksDB?). > And also a Ratis enabled cluster to HA cluster, I think this can work, if the config has HA style no issues, else if not HA style again they need to use omServiceIdDefault. (No need of download snapshot out side ratis, as this cluster from begining it is ratis enabled) Can we have a follow-up Jira to give have some documentation on how to handle the above cases? Sure. I will open a follow up Jira for this. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
