linxiaokun528 commented on PR #4577: URL: https://github.com/apache/solr/pull/4577#issuecomment-4931356806
> Thank you for submitting this. I began taking a look as it is an aggravating factor in another bug I've mentioned this in. I don't love that we are trying to reverse engineer granular zk transition states from coarser curator events though I personally don't have any better ideas yet. Something about this feels like we are fighting the framework. Though I definitely agree that doing this leader churn on every disconnect is totally unsutainable. It may very well be that solr is ill-equipped to take advantage of curator in its current form. @kotman12 Oops, clicked 'Close' by mistake! Reopening. Here are my thoughts on your previous comments... Anyway here is my take: I don't necessarily view this as fighting the framework; rather, it’s more of a necessary enhancement/workaround to bridge the framework's operational gaps. As a side note, I did identify the "Zombie Leader" problem as well. I full plan to tackle that root cause in a subsequent PR after this one. While this current PR doesn’t completely eliminate the issue, it drastically reduces the likelihood of triggering it in our environment. Given that Solr 10 is currently causing severe stability issues in our environment, could we consider merging this PR as a tactical mitigation for now? Investigating a brand-new approach or fully fixing the Zombie Leader bug will undoubtedly be a long-term effort that could span multiple Solr releases, whereas we desperately need relief right now. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
