The proposal looks good to me. I noticed the PR will change the existing RangeEntryCacheImpl to ship this improvement. Is it better to have a new Class name and make it configurable? The existing RangeEntryCacheImpl has been used for many years, we'd better consider a way to roll out the improvement to the hot path smoothly. We can keep the new implementation enabled by default, but allow users to change back to the existing solution without rolling back the entire cluster.
Best, Penghui On Thu, Jul 24, 2025 at 11:47 PM Matteo Merli <[email protected]> wrote: > +1 binding > -- > Matteo Merli > <[email protected]> > > > On Fri, Jul 25, 2025 at 3:37 AM Yunze Xu <[email protected]> wrote: > > > +1 (binding) > > > > Thanks, > > Yunze > > > > On Thu, Jul 24, 2025 at 9:47 PM Lari Hotari <[email protected]> wrote: > > > > > > Hi, > > > > > > I'd like to start the voting thread for "PIP-430: Pulsar Broker cache > > > improvements: refactoring eviction and adding a new cache strategy > > > based on expected read count". > > > > > > Proposal PR: https://github.com/apache/pulsar/pull/24444 > > > Rendered PIP document: > > > https://github.com/lhotari/pulsar/blob/lh-pip-430/pip/pip-430.md > > > Discussion thread: > > > https://lists.apache.org/thread/o1ozbg468kxfd38pxk2ppzsstdnxnok2 > > > > > > Please review the PIP document and vote! > > > The vote will stay open for at least 48 hours. > > > > > > Thanks, > > > > > > Lari > > >
