open https://issues.apache.org/jira/browse/KAFKA-19237
Chia-Ping Tsai <chia7...@gmail.com> 於 2025年5月3日 週六 下午11:35寫道： > hi Kamal > > Thanks for the response. I will file a jira and KIP for it. > > Best, > Chia-Ping > > > Kamal Chandraprakash <kamal.chandraprak...@gmail.com> 於 2025年5月3日 > 晚上10:54 寫道： > > > > 1. Yes, it was considered during the implementation of KIP-950 but the > > config was not proposed in the KIP and > > the KIP/Feature freeze date was passed by that time. > > > > 2. Having the RLMFollowerTask thread dynamically configurable is a useful > > feature. > > > > I'm +1 on adding the config to maintain consistency. We need to propose > the > > KIP for it. > > > > PR: https://github.com/apache/kafka/pull/16502 > > > > Thanks, > > Kamal > > > > > > > >> On Sat, May 3, 2025 at 2:51 PM Chia-Ping Tsai <chia7...@gmail.com> > wrote: > >> > >> hi Kamal > >> > >> Thanks for your response. I have two minor questions. > >> > >> 1. Have we considered adding a new config for RLMFolloweTask, such as > >> remote.log.manager.follower.thread.pool.size, for consistency? > >> > >> 2. Additionally, with the specific config, the size of RLMFolloweTask > can > >> be dynamically configured > >> > >> WDYT? > >> > >> Best, > >> Chia-Ping > >> > >> > >> > >>> Kamal Chandraprakash <kamal.chandraprak...@gmail.com> 於 2025年5月3日 > >> 下午2:29 寫道： > >>> > >>> Hi Chia-Ping, > >>> > >>> Thanks for following up on this! > >>> > >>> The `remote.log.manager.thread.pool.size` was initially planned to be > >>> deprecated as part of KIP-950 > >>> < > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-950%3A++Tiered+Storage+Disablement > >>> , > >>> > >>> then it was repurposed to use it as followerThreadPool in KIP-1030 > >>> < > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1030%3A+Change+constraints+and+default+values+for+various+configurations#KIP1030:Changeconstraintsanddefaultvaluesforvariousconfigurations-#9#10#11Config(s):remote.log.manager.copier.thread.pool.size/remote.log.manager.expiration.thread.pool.size/remote.log.manager.thread.pool.size > <https://cwiki.apache.org/confluence/display/KAFKA/KIP-1030%3A+Change+constraints+and+default+values+for+various+configurations#KIP1030:Changeconstraintsanddefaultvaluesforvariousconfigurations-%239%2310%2311Config(s):remote.log.manager.copier.thread.pool.size/remote.log.manager.expiration.thread.pool.size/remote.log.manager.thread.pool.size> > >> < > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1030%3A+Change+constraints+and+default+values+for+various+configurations#KIP1030:Changeconstraintsanddefaultvaluesforvariousconfigurations-%239%2310%2311Config(s):remote.log.manager.copier.thread.pool.size/remote.log.manager.expiration.thread.pool.size/remote.log.manager.thread.pool.size > > > >>> > >>> . > >>> > >>> PR: https://github.com/apache/kafka/pull/18137 > >>> > >>> -- > >>> Kamal > >>> > >>> > >>>> On Sat, May 3, 2025 at 11:09 AM Chia-Ping Tsai <chia7...@apache.org> > >> wrote: > >>>> > >>>> hi Kamal > >>>> > >>>> Apologies for posting on this older thread. I have a question > regarding > >>>> the configuration parameter remote.log.manager.thread.pool.size. > >>>> > >>>> Is there a Jira ticket associated with the deprecation of this > setting? > >>>> Additionally, is there a replacement configuration available for > >>>> followerThreadPool? > >>>> > >>>> thanks, > >>>> Chia-Ping > >>>> > >>>>> On 2024/11/07 16:07:36 Kamal Chandraprakash wrote: > >>>>> Hi Satish, > >>>>> > >>>>> Thanks for the review! Yes, we won't be enabling the deprecated > config > >> as > >>>>> dynamic. > >>>>> Removed the remote.log.manager.thread.pool.size config from the KIP. > >>>>> > >>>>> Currently, the config is not marked as deprecated which was > >>>>> already proposed in > >>>>> KIP-950, we can mark that config as deprecated in the source code. > >>>>> > >>>>> Thanks, > >>>>> Kamal > >>>>> > >>>>> On Thu, Nov 7, 2024 at 9:23 PM Satish Duggana < > >> satish.dugg...@gmail.com> > >>>>> wrote: > >>>>> > >>>>>> Thanks Kamal for the KIP. This is useful for dynamically changing > the > >>>>>> thread pool configurations, especially in production environments. > We > >>>>>> can skip remote.log.manager.thread.pool.size as it is already > >>>>>> deprecated, and remote.log.manager.copier.thread.pool.size, > >>>>>> remote.log.manager.expiration.thread.pool.size are derived from > that. > >>>>>> > >>>>>> ~Satish. > >>>>>> > >>>>>> > >>>>>> On Thu, 7 Nov 2024 at 10:07, Kamal Chandraprakash > >>>>>> <kamal.chandraprak...@gmail.com> wrote: > >>>>>>> > >>>>>>> Hi all, > >>>>>>> > >>>>>>> If there are no more comments, then I'll start a voting thread as > the > >>>>>>> change is minor. > >>>>>>> > >>>>>>> -- > >>>>>>> Kamal > >>>>>>> > >>>>>>> On Thu, Nov 7, 2024 at 9:00 AM Kamal Chandraprakash < > >>>>>>> kamal.chandraprak...@gmail.com> wrote: > >>>>>>> > >>>>>>>> Hi Federico, > >>>>>>>> > >>>>>>>> Updated the KIP by replacing the `isInitialized` to `isReady` in > >>>> the > >>>>>> KIP. > >>>>>>>> > >>>>>>>> > >>>>>>>> > >>>>>>>> On Wed, Nov 6, 2024 at 12:47 PM Federico Valeri < > >>>> fedeval...@gmail.com> > >>>>>>>> wrote: > >>>>>>>> > >>>>>>>>> Thanks Kamal, LGTM, but you should replace all instances of > >>>>>>>>> isInitialized to isReady in the rest of the KIP. > >>>>>>>>> > >>>>>>>>> On Wed, Nov 6, 2024 at 5:22 AM Kamal Chandraprakash > >>>>>>>>> <kamal.chandraprak...@gmail.com> wrote: > >>>>>>>>>> > >>>>>>>>>> Hi Federico, > >>>>>>>>>> > >>>>>>>>>> Thanks for the review! > >>>>>>>>>> > >>>>>>>>>> 1. Changed the API name to `isReady` > >>>>>>>>>> 2. Added an example of stacktrace in the KIP. > >>>>>>>>>> > >>>>>>>>>> PTAL. > >>>>>>>>>> > >>>>>>>>>> Thanks, > >>>>>>>>>> Kamal > >>>>>>>>>> > >>>>>>>>>> On Mon, Nov 4, 2024 at 2:37 PM Federico Valeri < > >>>>>> fedeval...@gmail.com> > >>>>>>>>> wrote: > >>>>>>>>>> > >>>>>>>>>>> Hi Kamal, these changes make sense to me. Thanks. > >>>>>>>>>>> > >>>>>>>>>>> In this case, I wonder if "isReady" could be a better name, > >>>>>> instead of > >>>>>>>>>>> "isInitialized". Wdyt? > >>>>>>>>>>> > >>>>>>>>>>> Could you please add an example of the stack trace that the > >>>> RLMM > >>>>>> can > >>>>>>>>>>> raise during the initialization phase? > >>>>>>>>>>> > >>>>>>>>>>> On Sun, Nov 3, 2024 at 4:50 PM Kamal Chandraprakash > >>>>>>>>>>> <kamal.chandraprak...@gmail.com> wrote: > >>>>>>>>>>>> > >>>>>>>>>>>> Hi all, > >>>>>>>>>>>> > >>>>>>>>>>>> I would like to start a discussion thread on KIP-1105 > >>>>>>>>>>>> < > >>>>>>>>>>> > >>>>>>>>> > >>>>>> > >>>> > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1105%3A+Make+remote+log+manager+thread-pool+configs+dynamic > >>>>>>>>>>>> . > >>>>>>>>>>>> This KIP is about > >>>>>>>>>>>> > >>>>>>>>>>>> 1. Configuring the thread-pool used by the remote-log > >>>> manager > >>>>>>>>> dynamically > >>>>>>>>>>>> and > >>>>>>>>>>>> 2. Graceful handling of remote-log components during server > >>>>>> startup. > >>>>>>>>>>>> > >>>>>>>>>>>> > >>>>>>>>>>> > >>>>>>>>> > >>>>>> > >>>> > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1105%3A+Make+remote+log+manager+thread-pool+configs+dynamic > >>>>>>>>>>>> > >>>>>>>>>>>> Please take a look and suggest your thoughts. > >>>>>>>>>>>> > >>>>>>>>>>>> Thanks, > >>>>>>>>>>>> Kamal > >>>>>>>>>>> > >>>>>>>>> > >>>>>>>> > >>>>>> > >>>>> > >>>> > >> >
