The cache size is 8000，allow the cache map to grow up to 8000 entries and then delete the eldest entry each time a new entry is added, maintaining a steady state of 8000 entries.The addition logic of the cache is to write to the disk asynchronously or synchronously, and then add it to the cache; the read logic of the cache is to read the cache first, and then read the disk if it does not hit; if it hits, read the cache and delete it; if the broker reboot, all the cache will be lost.
Xinyu Zhou <[email protected]> 于2022年10月11日周二 15:51写道： > Great enhancement for transactional messages! > > Could you please have more details about how the TransactionCacheManager > handles the abnormal reboot of the broker? The reboot may fail the > commit/rollback operations. > > Regards > > On Tue, Oct 11, 2022 at 2:54 PM fuyou <[email protected]> wrote: > > > good proposal . > > > > > > > > tihong ruan <[email protected]> 于2022年10月11日周二 11:53写道： > > > > > Hi, RocketMQ Community, > > > Apache RocketMQ provides a distributed transaction feature that is > > similar > > > to X/Open XA, ensuring transaction consistency. We want to do some > > > optimizations on this module. > > > > > > We have written the proposal and you can see it by the link below: > > > https://shimo.im/docs/2wAlXDnwoPsjybAP/ 「RIP 50 RocketMQ Transaction > > > Message Improvement」 > > > > > > Please reply to this email if you have any questions > > > > > > > > > -- > > ============================================= > > > > fuyou001 > > Best Regards > > >
