eolivelli commented on pull request #14176: URL: https://github.com/apache/pulsar/pull/14176#issuecomment-1033491553
@hadesy thanks for your detailed explanation. This patch will mitigate the problem, and we could commit it in order to take time before eventually finding other problems. The downside is that we will hide other problems if we commit this patch. we should prevent any blocking call to the metadata service on the metadata executor (IIUC here we are no more talking about the ZK client thread, but our own Metadata listener thread). We could add some check about the fact that the "current thread" is not the "metadata-store" thread in blocking calls to the MetadataStore. IIRC in Netty they are doing so in order to prevent this kind of errors. Overall I am leaning towards accepting this patch as we cannot know all of the problems like the one you found here (and we should fix the specific case) Probably a broader discussion on dev@ mailing is better, as not everybody is following GH notifications about PRs and this topic is very important -- 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]
