回复:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Hi, team I updated the KIP-571 since we took a slightly different implementation in the PR https://github.com/apache/kafka/pull/8589, basically: In RemoveMembersFromConsumerGroupOptions, leveraging empty members rather than introducing a new field to imply the removeAll scenario. Please let me know if you have any concerns, thanks a lot! Feyman -- 发件人:feyman2009 发送时间:2020年4月13日(星期一) 08:47 收件人:dev 主 题:回复:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Thanks , John and Guochang! -- 发件人:Guozhang Wang 发送时间:2020年4月11日(星期六) 03:07 收件人:dev 主 题:Re: 回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Thanks Feyman, I've looked at the update that you incorporated from Matthias and that LGTM too. I'm still +1 :) Guozhang On Fri, Apr 10, 2020 at 11:18 AM John Roesler wrote: > Hey Feyman, > > Just to remove any ambiguity, I've been casually following the discussion, > I've just looked at the KIP document again, and I'm still +1 (binding). > > Thanks, > -John > > On Fri, Apr 10, 2020, at 01:44, feyman2009 wrote: > > Hi, all > > KIP-571 has already collected 4 bind +1 (John, Guochang, Bill, > > Matthias) and 3 non-binding +1(Boyang, Sophie), I will mark it as > > approved and create a PR shortly. > > Thanks! > > > > Feyman > > -- > > 发件人:feyman2009 > > 发送时间:2020年4月8日(星期三) 14:21 > > 收件人:dev ; Boyang Chen > > 主 题:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hi Boyang, > > Thanks for reminding me of that! > > I'm not sure about the convention, I thought it would need to > > re-collect votes if the KIP has changed~ > > Let's leave the vote thread here for 2 days, if no objection, I > > will take it as approved and update the PR accordingly. > > > > Thanks! > > Feyman > > > > > > > > ------------------ > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 12:42 > > 收件人:dev ; feyman2009 > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > You should already get enough votes if I'm counting correctly > > (Guozhang, John, Matthias) > > On Tue, Apr 7, 2020 at 6:59 PM feyman2009 > > wrote: > > Hi, Boyang > > I think Matthias's proposal makes sense, but we can use the admin > > tool for this scenario as Boyang mentioned or follow up later, so I > > prefer to keep this KIP unchanged to minimize the scope. > > Calling for vote ~ > > > > Thanks! > > Feyman > > > > -- > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 02:15 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hey Feyman, > > > > I think Matthias' suggestion is optional, and we could just use admin > tool > > to remove single static members as well. > > > > Boyang > > > > On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax > wrote: > > > > > > Would you mind to elaborate why we still need that > > > > > > Sure. > > > > > > For static memership, the session timeout it usually set quite high. > > > This make scaling in an application tricky: if you shut down one > > > instance, no rebalance would happen until `session.timeout.ms` hits. > > > This is specific to Kafka Streams, because when a Kafka Stream > > client is > > > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > > > corresponding partitions would not be processed for a long time and > > > thus, fall back. > > > > > > Given that each instance will have a unique `instance.id` provided by > > > the user, we could allow users to remove the instance they want to > > > decommission from the consumer group without the need to wait for > > > `session.timeout.ms`. > > > > > > Hence, it's not an application reset scenario for which one wants to > > > remove all members, but a scaling-in scenario. For dynamic > > membership, > > > this issue usually does not occur because the `session.timeout.ms` is > > > set to a fairly low value and a rebalance would happen quick
回复:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Thanks , John and Guochang! -- 发件人:Guozhang Wang 发送时间:2020年4月11日(星期六) 03:07 收件人:dev 主 题:Re: 回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Thanks Feyman, I've looked at the update that you incorporated from Matthias and that LGTM too. I'm still +1 :) Guozhang On Fri, Apr 10, 2020 at 11:18 AM John Roesler wrote: > Hey Feyman, > > Just to remove any ambiguity, I've been casually following the discussion, > I've just looked at the KIP document again, and I'm still +1 (binding). > > Thanks, > -John > > On Fri, Apr 10, 2020, at 01:44, feyman2009 wrote: > > Hi, all > > KIP-571 has already collected 4 bind +1 (John, Guochang, Bill, > > Matthias) and 3 non-binding +1(Boyang, Sophie), I will mark it as > > approved and create a PR shortly. > > Thanks! > > > > Feyman > > -- > > 发件人:feyman2009 > > 发送时间:2020年4月8日(星期三) 14:21 > > 收件人:dev ; Boyang Chen > > 主 题:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hi Boyang, > > Thanks for reminding me of that! > > I'm not sure about the convention, I thought it would need to > > re-collect votes if the KIP has changed~ > > Let's leave the vote thread here for 2 days, if no objection, I > > will take it as approved and update the PR accordingly. > > > > Thanks! > > Feyman > > > > > > > > ------------------ > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 12:42 > > 收件人:dev ; feyman2009 > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > You should already get enough votes if I'm counting correctly > > (Guozhang, John, Matthias) > > On Tue, Apr 7, 2020 at 6:59 PM feyman2009 > > wrote: > > Hi, Boyang > > I think Matthias's proposal makes sense, but we can use the admin > > tool for this scenario as Boyang mentioned or follow up later, so I > > prefer to keep this KIP unchanged to minimize the scope. > > Calling for vote ~ > > > > Thanks! > > Feyman > > > > -- > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 02:15 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hey Feyman, > > > > I think Matthias' suggestion is optional, and we could just use admin > tool > > to remove single static members as well. > > > > Boyang > > > > On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax > wrote: > > > > > > Would you mind to elaborate why we still need that > > > > > > Sure. > > > > > > For static memership, the session timeout it usually set quite high. > > > This make scaling in an application tricky: if you shut down one > > > instance, no rebalance would happen until `session.timeout.ms` hits. > > > This is specific to Kafka Streams, because when a Kafka Stream > > client is > > > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > > > corresponding partitions would not be processed for a long time and > > > thus, fall back. > > > > > > Given that each instance will have a unique `instance.id` provided by > > > the user, we could allow users to remove the instance they want to > > > decommission from the consumer group without the need to wait for > > > `session.timeout.ms`. > > > > > > Hence, it's not an application reset scenario for which one wants to > > > remove all members, but a scaling-in scenario. For dynamic > > membership, > > > this issue usually does not occur because the `session.timeout.ms` is > > > set to a fairly low value and a rebalance would happen quickly after > > an > > > instance is decommissioned. > > > > > > Does this make sense? > > > > > > As said before, we may or may not include this in this KIP. It's up > > to > > > you if you want to address it or not. > > > > > > > > > -Matthias > > > > > > > > > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > > > Hi, Matthias > > > > Thanks a lot! > > > > So you do not plan so support removing a _single stat
Re: 回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Thanks Feyman, I've looked at the update that you incorporated from Matthias and that LGTM too. I'm still +1 :) Guozhang On Fri, Apr 10, 2020 at 11:18 AM John Roesler wrote: > Hey Feyman, > > Just to remove any ambiguity, I've been casually following the discussion, > I've just looked at the KIP document again, and I'm still +1 (binding). > > Thanks, > -John > > On Fri, Apr 10, 2020, at 01:44, feyman2009 wrote: > > Hi, all > > KIP-571 has already collected 4 bind +1 (John, Guochang, Bill, > > Matthias) and 3 non-binding +1(Boyang, Sophie), I will mark it as > > approved and create a PR shortly. > > Thanks! > > > > Feyman > > -- > > 发件人:feyman2009 > > 发送时间:2020年4月8日(星期三) 14:21 > > 收件人:dev ; Boyang Chen > > 主 题:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hi Boyang, > > Thanks for reminding me of that! > > I'm not sure about the convention, I thought it would need to > > re-collect votes if the KIP has changed~ > > Let's leave the vote thread here for 2 days, if no objection, I > > will take it as approved and update the PR accordingly. > > > > Thanks! > > Feyman > > > > > > > > ------------------ > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 12:42 > > 收件人:dev ; feyman2009 > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > You should already get enough votes if I'm counting correctly > > (Guozhang, John, Matthias) > > On Tue, Apr 7, 2020 at 6:59 PM feyman2009 > > wrote: > > Hi, Boyang > > I think Matthias's proposal makes sense, but we can use the admin > > tool for this scenario as Boyang mentioned or follow up later, so I > > prefer to keep this KIP unchanged to minimize the scope. > > Calling for vote ~ > > > > Thanks! > > Feyman > > > > -- > > 发件人:Boyang Chen > > 发送时间:2020年4月8日(星期三) 02:15 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > > members in StreamsResetter > > > > Hey Feyman, > > > > I think Matthias' suggestion is optional, and we could just use admin > tool > > to remove single static members as well. > > > > Boyang > > > > On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax > wrote: > > > > > > Would you mind to elaborate why we still need that > > > > > > Sure. > > > > > > For static memership, the session timeout it usually set quite high. > > > This make scaling in an application tricky: if you shut down one > > > instance, no rebalance would happen until `session.timeout.ms` hits. > > > This is specific to Kafka Streams, because when a Kafka Stream > > client is > > > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > > > corresponding partitions would not be processed for a long time and > > > thus, fall back. > > > > > > Given that each instance will have a unique `instance.id` provided by > > > the user, we could allow users to remove the instance they want to > > > decommission from the consumer group without the need to wait for > > > `session.timeout.ms`. > > > > > > Hence, it's not an application reset scenario for which one wants to > > > remove all members, but a scaling-in scenario. For dynamic > > membership, > > > this issue usually does not occur because the `session.timeout.ms` is > > > set to a fairly low value and a rebalance would happen quickly after > > an > > > instance is decommissioned. > > > > > > Does this make sense? > > > > > > As said before, we may or may not include this in this KIP. It's up > > to > > > you if you want to address it or not. > > > > > > > > > -Matthias > > > > > > > > > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > > > Hi, Matthias > > > > Thanks a lot! > > > > So you do not plan so support removing a _single static_ > > member via > > > `StreamsResetter`? > > > > => > > > > Would you mind to elaborate why we still need that if we > > are > > > able to batch remove active m
Re: 回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Hey Feyman, Just to remove any ambiguity, I've been casually following the discussion, I've just looked at the KIP document again, and I'm still +1 (binding). Thanks, -John On Fri, Apr 10, 2020, at 01:44, feyman2009 wrote: > Hi, all > KIP-571 has already collected 4 bind +1 (John, Guochang, Bill, > Matthias) and 3 non-binding +1(Boyang, Sophie), I will mark it as > approved and create a PR shortly. > Thanks! > > Feyman > -- > 发件人:feyman2009 > 发送时间:2020年4月8日(星期三) 14:21 > 收件人:dev ; Boyang Chen > 主 题:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > members in StreamsResetter > > Hi Boyang, > Thanks for reminding me of that! > I'm not sure about the convention, I thought it would need to > re-collect votes if the KIP has changed~ > Let's leave the vote thread here for 2 days, if no objection, I > will take it as approved and update the PR accordingly. > > Thanks! > Feyman > > > > -- > 发件人:Boyang Chen > 发送时间:2020年4月8日(星期三) 12:42 > 收件人:dev ; feyman2009 > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > members in StreamsResetter > > You should already get enough votes if I'm counting correctly > (Guozhang, John, Matthias) > On Tue, Apr 7, 2020 at 6:59 PM feyman2009 > wrote: > Hi, Boyang > I think Matthias's proposal makes sense, but we can use the admin > tool for this scenario as Boyang mentioned or follow up later, so I > prefer to keep this KIP unchanged to minimize the scope. > Calling for vote ~ > > Thanks! > Feyman > > -------------------------------------- > 发件人:Boyang Chen > 发送时间:2020年4月8日(星期三) 02:15 > 收件人:dev > 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > members in StreamsResetter > > Hey Feyman, > > I think Matthias' suggestion is optional, and we could just use admin tool > to remove single static members as well. > > Boyang > > On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax wrote: > > > > Would you mind to elaborate why we still need that > > > > Sure. > > > > For static memership, the session timeout it usually set quite high. > > This make scaling in an application tricky: if you shut down one > > instance, no rebalance would happen until `session.timeout.ms` hits. > > This is specific to Kafka Streams, because when a Kafka Stream > client is > > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > > corresponding partitions would not be processed for a long time and > > thus, fall back. > > > > Given that each instance will have a unique `instance.id` provided by > > the user, we could allow users to remove the instance they want to > > decommission from the consumer group without the need to wait for > > `session.timeout.ms`. > > > > Hence, it's not an application reset scenario for which one wants to > > remove all members, but a scaling-in scenario. For dynamic > membership, > > this issue usually does not occur because the `session.timeout.ms` is > > set to a fairly low value and a rebalance would happen quickly after > an > > instance is decommissioned. > > > > Does this make sense? > > > > As said before, we may or may not include this in this KIP. It's up > to > > you if you want to address it or not. > > > > > > -Matthias > > > > > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > > Hi, Matthias > > > Thanks a lot! > > > So you do not plan so support removing a _single static_ > member via > > `StreamsResetter`? > > > => > > > Would you mind to elaborate why we still need that if we > are > > able to batch remove active members with adminClient? > > > > > > Thanks! > > > > > > Feyman > > > -- > > > 发件人:Matthias J. Sax > > > 发送时间:2020年4月7日(星期二) 08:25 > > > 收件人:dev > > > 主 题:Re: 回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove > members > > in StreamsResetter > > > > > > Overall LGTM. > > > > > > +1 (binding) > > > > > > So you do not plan so support removing a _single static_ member via > > > `StreamsResetter`? We can of course still add this as a follow up > but it > > > might be nice to just add
回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Hi, all KIP-571 has already collected 4 bind +1 (John, Guochang, Bill, Matthias) and 3 non-binding +1(Boyang, Sophie), I will mark it as approved and create a PR shortly. Thanks! Feyman -- 发件人:feyman2009 发送时间:2020年4月8日(星期三) 14:21 收件人:dev ; Boyang Chen 主 题:回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Hi Boyang, Thanks for reminding me of that! I'm not sure about the convention, I thought it would need to re-collect votes if the KIP has changed~ Let's leave the vote thread here for 2 days, if no objection, I will take it as approved and update the PR accordingly. Thanks! Feyman -- 发件人:Boyang Chen 发送时间:2020年4月8日(星期三) 12:42 收件人:dev ; feyman2009 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter You should already get enough votes if I'm counting correctly (Guozhang, John, Matthias) On Tue, Apr 7, 2020 at 6:59 PM feyman2009 wrote: Hi, Boyang I think Matthias's proposal makes sense, but we can use the admin tool for this scenario as Boyang mentioned or follow up later, so I prefer to keep this KIP unchanged to minimize the scope. Calling for vote ~ Thanks! Feyman -- 发件人:Boyang Chen 发送时间:2020年4月8日(星期三) 02:15 收件人:dev 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Hey Feyman, I think Matthias' suggestion is optional, and we could just use admin tool to remove single static members as well. Boyang On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax wrote: > > Would you mind to elaborate why we still need that > > Sure. > > For static memership, the session timeout it usually set quite high. > This make scaling in an application tricky: if you shut down one > instance, no rebalance would happen until `session.timeout.ms` hits. > This is specific to Kafka Streams, because when a Kafka Stream client is > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > corresponding partitions would not be processed for a long time and > thus, fall back. > > Given that each instance will have a unique `instance.id` provided by > the user, we could allow users to remove the instance they want to > decommission from the consumer group without the need to wait for > `session.timeout.ms`. > > Hence, it's not an application reset scenario for which one wants to > remove all members, but a scaling-in scenario. For dynamic membership, > this issue usually does not occur because the `session.timeout.ms` is > set to a fairly low value and a rebalance would happen quickly after an > instance is decommissioned. > > Does this make sense? > > As said before, we may or may not include this in this KIP. It's up to > you if you want to address it or not. > > > -Matthias > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > Hi, Matthias > > Thanks a lot! > > So you do not plan so support removing a _single static_ member via > `StreamsResetter`? > > => > > Would you mind to elaborate why we still need that if we are > able to batch remove active members with adminClient? > > > > Thanks! > > > > Feyman > > -------------------------- > > 发件人:Matthias J. Sax > > 发送时间:2020年4月7日(星期二) 08:25 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members > in StreamsResetter > > > > Overall LGTM. > > > > +1 (binding) > > > > So you do not plan so support removing a _single static_ member via > > `StreamsResetter`? We can of course still add this as a follow up but it > > might be nice to just add it to this KIP right away. Up to you if you > > want to include it or not. > > > > > > -Matthias > > > > > > > > On 3/30/20 8:16 AM, feyman2009 wrote: > >> Hi, Boyang > >> Thanks a lot, that make sense, we should not expose member.id in > the MemberToRemove anymore, I have fixed it in the KIP. > >> > >> > >> Feyman > >> -- > >> 发件人:Boyang Chen > >> 发送时间:2020年3月29日(星期日) 01:45 > >> 收件人:dev ; feyman2009 > >> 主 题:Re: 回复:回复:回复:[Vote] KIP-571: Add option to force remove members in > StreamsResetter > >> > >> Hey Feyman, > >> > >> thanks for the update. I ass
回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Hi Boyang, Thanks for reminding me of that! I'm not sure about the convention, I thought it would need to re-collect votes if the KIP has changed~ Let's leave the vote thread here for 2 days, if no objection, I will take it as approved and update the PR accordingly. Thanks! Feyman -- 发件人:Boyang Chen 发送时间:2020年4月8日(星期三) 12:42 收件人:dev ; feyman2009 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter You should already get enough votes if I'm counting correctly (Guozhang, John, Matthias) On Tue, Apr 7, 2020 at 6:59 PM feyman2009 wrote: Hi, Boyang I think Matthias's proposal makes sense, but we can use the admin tool for this scenario as Boyang mentioned or follow up later, so I prefer to keep this KIP unchanged to minimize the scope. Calling for vote ~ Thanks! Feyman -- 发件人:Boyang Chen 发送时间:2020年4月8日(星期三) 02:15 收件人:dev 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Hey Feyman, I think Matthias' suggestion is optional, and we could just use admin tool to remove single static members as well. Boyang On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax wrote: > > Would you mind to elaborate why we still need that > > Sure. > > For static memership, the session timeout it usually set quite high. > This make scaling in an application tricky: if you shut down one > instance, no rebalance would happen until `session.timeout.ms` hits. > This is specific to Kafka Streams, because when a Kafka Stream client is > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > corresponding partitions would not be processed for a long time and > thus, fall back. > > Given that each instance will have a unique `instance.id` provided by > the user, we could allow users to remove the instance they want to > decommission from the consumer group without the need to wait for > `session.timeout.ms`. > > Hence, it's not an application reset scenario for which one wants to > remove all members, but a scaling-in scenario. For dynamic membership, > this issue usually does not occur because the `session.timeout.ms` is > set to a fairly low value and a rebalance would happen quickly after an > instance is decommissioned. > > Does this make sense? > > As said before, we may or may not include this in this KIP. It's up to > you if you want to address it or not. > > > -Matthias > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > Hi, Matthias > > Thanks a lot! > > So you do not plan so support removing a _single static_ member via > `StreamsResetter`? > > => > > Would you mind to elaborate why we still need that if we are > able to batch remove active members with adminClient? > > > > Thanks! > > > > Feyman > > -------------------------- > > 发件人:Matthias J. Sax > > 发送时间:2020年4月7日(星期二) 08:25 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members > in StreamsResetter > > > > Overall LGTM. > > > > +1 (binding) > > > > So you do not plan so support removing a _single static_ member via > > `StreamsResetter`? We can of course still add this as a follow up but it > > might be nice to just add it to this KIP right away. Up to you if you > > want to include it or not. > > > > > > -Matthias > > > > > > > > On 3/30/20 8:16 AM, feyman2009 wrote: > >> Hi, Boyang > >> Thanks a lot, that make sense, we should not expose member.id in > the MemberToRemove anymore, I have fixed it in the KIP. > >> > >> > >> Feyman > >> -- > >> 发件人:Boyang Chen > >> 发送时间:2020年3月29日(星期日) 01:45 > >> 收件人:dev ; feyman2009 > >> 主 题:Re: 回复:回复:回复:[Vote] KIP-571: Add option to force remove members in > StreamsResetter > >> > >> Hey Feyman, > >> > >> thanks for the update. I assume we would rely entirely on the internal > changes for `removeMemberFromGroup` by sending a DescribeGroup request > first. With that in mind, I don't think we need to add member.id to > MemberToRemove anymore, as it is only facing public where users will only > configure group.instance.id? > >> On Fri, Mar 27, 2020 at 5:04 PM feyman2009 > wrote: > >> Bump, can anyone kindly ta
回复:回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter
Hi, Boyang I think Matthias's proposal makes sense, but we can use the admin tool for this scenario as Boyang mentioned or follow up later, so I prefer to keep this KIP unchanged to minimize the scope. Calling for vote ~ Thanks! Feyman -- 发件人:Boyang Chen 发送时间:2020年4月8日(星期三) 02:15 收件人:dev 主 题:Re: 回复:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in StreamsResetter Hey Feyman, I think Matthias' suggestion is optional, and we could just use admin tool to remove single static members as well. Boyang On Tue, Apr 7, 2020 at 11:00 AM Matthias J. Sax wrote: > > Would you mind to elaborate why we still need that > > Sure. > > For static memership, the session timeout it usually set quite high. > This make scaling in an application tricky: if you shut down one > instance, no rebalance would happen until `session.timeout.ms` hits. > This is specific to Kafka Streams, because when a Kafka Stream client is > closed, it does _not_ send a `LeaveGroupRequest`. Hence, the > corresponding partitions would not be processed for a long time and > thus, fall back. > > Given that each instance will have a unique `instance.id` provided by > the user, we could allow users to remove the instance they want to > decommission from the consumer group without the need to wait for > `session.timeout.ms`. > > Hence, it's not an application reset scenario for which one wants to > remove all members, but a scaling-in scenario. For dynamic membership, > this issue usually does not occur because the `session.timeout.ms` is > set to a fairly low value and a rebalance would happen quickly after an > instance is decommissioned. > > Does this make sense? > > As said before, we may or may not include this in this KIP. It's up to > you if you want to address it or not. > > > -Matthias > > > > On 4/7/20 7:12 AM, feyman2009 wrote: > > Hi, Matthias > > Thanks a lot! > > So you do not plan so support removing a _single static_ member via > `StreamsResetter`? > > => > > Would you mind to elaborate why we still need that if we are > able to batch remove active members with adminClient? > > > > Thanks! > > > > Feyman > > ---------------------- > > 发件人:Matthias J. Sax > > 发送时间:2020年4月7日(星期二) 08:25 > > 收件人:dev > > 主 题:Re: 回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members > in StreamsResetter > > > > Overall LGTM. > > > > +1 (binding) > > > > So you do not plan so support removing a _single static_ member via > > `StreamsResetter`? We can of course still add this as a follow up but it > > might be nice to just add it to this KIP right away. Up to you if you > > want to include it or not. > > > > > > -Matthias > > > > > > > > On 3/30/20 8:16 AM, feyman2009 wrote: > >> Hi, Boyang > >> Thanks a lot, that make sense, we should not expose member.id in > the MemberToRemove anymore, I have fixed it in the KIP. > >> > >> > >> Feyman > >> -- > >> 发件人:Boyang Chen > >> 发送时间:2020年3月29日(星期日) 01:45 > >> 收件人:dev ; feyman2009 > >> 主 题:Re: 回复:回复:回复:[Vote] KIP-571: Add option to force remove members in > StreamsResetter > >> > >> Hey Feyman, > >> > >> thanks for the update. I assume we would rely entirely on the internal > changes for `removeMemberFromGroup` by sending a DescribeGroup request > first. With that in mind, I don't think we need to add member.id to > MemberToRemove anymore, as it is only facing public where users will only > configure group.instance.id? > >> On Fri, Mar 27, 2020 at 5:04 PM feyman2009 > wrote: > >> Bump, can anyone kindly take a look at the updated KIP-571? Thanks! > >> > >> > >> -- > >> 发件人:feyman2009 > >> 发送时间:2020年3月23日(星期一) 08:51 > >> 收件人:dev > >> 主 题:回复:回复:回复:回复:[Vote] KIP-571: Add option to force remove members in > StreamsResetter > >> > >> Hi, team > >> I have updated the KIP-571 according to our latest discussion > results, would you mind to take a look? Thanks! > >> > >> Feyman > >> > >> > >> -- > >> 发件人:Boyang Chen > >> 发送时间:2020年3月19日(星期四) 13:41 > >> 收件人:dev ; feyman2009 > >> 主 题:Re: 回复:回复:回复:[Vote] KIP-571