Hi Jason, Thanks for your feedback.
If we want to keep the per-partition expiration timestamp, then the scope of the KIP will be significantly reduced. In other words, we will just be deleting offsets from cache as before but only if the group is in Empty state. This would not cause any negative backward compatibility concern, since offsets won't expire any earlier than before. They may be removed at the same time (if group is already Empty) or later (if group is not Empty yet) than before. If I'm not mistaken we no longer need to change the internal schema either - no need for keeping track of when group becomes Empty. Would this make sense? If so, I'll update the KIP with this reduced-scope proposal. Regarding manual offset removal I was thinking of synching up the new consumer group command with the old command in which a '--topic' option was supported with '--delete'. Regarding the other JIRA you referred to, sure, I'll add that in the KIP. Thanks. --Vahid From: Jason Gustafson <ja...@confluent.io> To: dev@kafka.apache.org Date: 02/28/2018 12:10 PM Subject: Re: [DISCUSS] KIP-211: Revise Expiration Semantics of Consumer Group Offsets Hey Vahid, Thanks for the response. Replies below: > 1. I think my suggestion in the KIP was more towards ignoring the client > provided values and use a large enough broker config value instead. It > seems the question comes down to whether we still want to honor the > `retention_time` field in the old requests. With the new request (as per > this KIP) the client would not be able to overwrite the broker retention > config. Your suggestion provides kind of a back door for the overwrite. > Also, since different offset commits associated with a group can > potentially use different `retention_time` values, it's probably > reasonable to use the maximum of all those values (including the broker > config) as the group offset retention. Mainly I wanted to ensure that we would be holding offsets at least as long as what was requested by older clients. If we hold it for longer, that's probably fine, but there may be application behavior which would break if offsets are expired earlier than expected. 2. If I'm not mistake you are referring to potential changes in > `GROUP_METADATA_VALUE_SCHEMA`. I saw this as an internal implementation > matter and frankly, have not fully thought about it, but I agree that it > needs to be updated to include either the timestamp the group becomes > `Empty` or maybe the expiration timestamp of the group. And perhaps, we > would not need to store per partition offset expiration timestamp anymore. > Is there a particular reason for your suggestion of storing the timestamp > the group becomes `Empty`, vs the expiration timestamp of the group? Although it is not exposed to clients, we still have to manage compatibility of the schema across versions, so I think we should include it in the KIP. The reason I was thinking of using the time that the group became Empty is that the configured timeout might change. I think my expectation as a user would be that a timeout change would also apply to existing groups, but I'm not sure if there are any reasons not to so. 3. To limit the scope of the KIP I would prefer to handle this matter > separately if it doesn't have to be addressed as part of this change. It > probably needs be addressed at some point and I'll mention it in the KIP > so we have it documented. Do you think my suggestion of manually removing > topic offsets from group (as an interim solution) is worth additional > discussion / implementation? I think manual removal of offsets for this case is a bit of a tough sell for usability. Did you imagine it happening automatically in the consumer through an API? I'm finding it increasingly frustrating that the generic group coordinator is limited in its decision making since it cannot see the subscription metadata. It is the same problem in Dong's KIP. I think I would suggest that, at a minimum, we leave the door open to enforcing offset expiration either 1) when the group becomes empty, and 2) when the corresponding partition is removed from the subscription. Perhaps that means we need to keep the individual offset expiration timestamp after all. Actually we would probably need it anyway to handle "simple" consumer groups which are always Empty. One additional note: I have seen recently a case where the offset cache caused an OOM on the broker. I looked into it and found that most of the cache was used for storing console consumer offsets. I know you had a patch before which turned off auto-commit when the groupId was generated by ConsoleConsumer. Maybe we could lump that change into this KIP? Thanks, Jason On Fri, Feb 23, 2018 at 4:08 PM, Vahid S Hashemian < vahidhashem...@us.ibm.com> wrote: > Hi Jason, > > Thanks a lot for reviewing the KIP. > > 1. I think my suggestion in the KIP was more towards ignoring the client > provided values and use a large enough broker config value instead. It > seems the question comes down to whether we still want to honor the > `retention_time` field in the old requests. With the new request (as per > this KIP) the client would not be able to overwrite the broker retention > config. Your suggestion provides kind of a back door for the overwrite. > Also, since different offset commits associated with a group can > potentially use different `retention_time` values, it's probably > reasonable to use the maximum of all those values (including the broker > config) as the group offset retention. > > 2. If I'm not mistake you are referring to potential changes in > `GROUP_METADATA_VALUE_SCHEMA`. I saw this as an internal implementation > matter and frankly, have not fully thought about it, but I agree that it > needs to be updated to include either the timestamp the group becomes > `Empty` or maybe the expiration timestamp of the group. And perhaps, we > would not need to store per partition offset expiration timestamp anymore. > Is there a particular reason for your suggestion of storing the timestamp > the group becomes `Empty`, vs the expiration timestamp of the group? > > 3. To limit the scope of the KIP I would prefer to handle this matter > separately if it doesn't have to be addressed as part of this change. It > probably needs be addressed at some point and I'll mention it in the KIP > so we have it documented. Do you think my suggestion of manually removing > topic offsets from group (as an interim solution) is worth additional > discussion / implementation? > > I'll wait for your feedback and clarification on the above items before > updating the KIP. > > Thanks. > --Vahid > > > > From: Jason Gustafson <ja...@confluent.io> > To: dev@kafka.apache.org > Date: 02/18/2018 01:16 PM > Subject: Re: [DISCUSS] KIP-211: Revise Expiration Semantics of > Consumer Group Offsets > > > > Hey Vahid, > > Sorry for the late response. The KIP looks good. A few comments: > > 1. I'm not quite sure I understand how you are handling old clients. It > sounds like you are saying that old clients need to change configuration? > I'd suggest 1) if an old client requests the default expiration, then we > use the updated behavior, and 2) if the old client requests a specific > expiration, we enforce it from the time the group becomes Empty. > > 2. Does this require a new version of the group metadata messsage format? > I > think we need to add a new field to indicate the time that the group state > changed to Empty. This will allow us to resume the expiration timer > correctly after a coordinator change. Alternatively, we could reset the > expiration timeout after every coordinator move, but it would be nice to > have a definite bound on offset expiration. > > 3. The question about removal of offsets for partitions which are no > longer > in use is interesting. At the moment, it's difficult for the coordinator > to > know that a partition is no longer being fetched because it is agnostic to > subscription state (the group coordinator is used for more than just > consumer groups). Even if we allow the coordinator to read subscription > state to tell which topics are no longer being consumed, we might need > some > additional bookkeeping to keep track of /when/ the consumer stopped > subscribing to a particular topic. Or maybe we can reset this expiration > timer after every coordinator change when the new coordinator reads the > offsets and group metadata? I am not sure how common this use case is and > whether it needs to be solved as part of this KIP. > > Thanks, > Jason > > > > On Thu, Feb 1, 2018 at 12:40 PM, Vahid S Hashemian < > vahidhashem...@us.ibm.com> wrote: > > > Thanks James for sharing that scenario. > > > > I agree it makes sense to be able to remove offsets for the topics that > > are no longer "active" in the group. > > I think it becomes important to determine what constitutes that a topic > is > > no longer active: If we use per-partition expiration we would manually > > choose a retention time that works for the particular scenario. > > > > That works, but since we are manually intervening and specify a > > per-partition retention, why not do the intervention in some other way: > > > > One alternative for this intervention, to favor the simplicity of the > > suggested protocol in the KIP, is to improve upon the just introduced > > DELETE_GROUPS API and allow for deletion of offsets of specific topics > in > > the group. This is what the old ZooKeeper based group management > supported > > anyway, and we would just be leveling the group deletion features of the > > Kafka-based group management with the ZooKeeper-based one. > > > > So, instead of deciding in advance when the offsets should be removed we > > would instantly remove them when we are sure that they are no longer > > needed. > > > > Let me know what you think. > > > > Thanks. > > --Vahid > > > > > > > > From: James Cheng <wushuja...@gmail.com> > > To: dev@kafka.apache.org > > Date: 02/01/2018 12:37 AM > > Subject: Re: [DISCUSS] KIP-211: Revise Expiration Semantics of > > Consumer Group Offsets > > > > > > > > Vahid, > > > > Under rejected alternatives, we had decided that we did NOT want to do > > per-partition expiration, and instead we wait until the entire group is > > empty and then (after the right time has passed) expire the entire group > > at once. > > > > I thought of one scenario that might benefit from per-partition > > expiration. > > > > Let's say I have topics A B C... Z. So, I have 26 topics, all of them > > single partition, so 26 partitions. Let's say I have mirrormaker > mirroring > > those 26 topics. The group will then have 26 committed offsets. > > > > Let's say I then change the whitelist on mirrormaker so that it only > > mirrors topic Z, but I keep the same consumer group name. (I imagine > that > > is a common thing to do?) > > > > With the proposed design for this KIP, the committed offsets for topics > A > > through Y will stay around as long as this mirroring group name exists. > > > > In the current implementation that already exists (prior to this KIP), I > > belive that committed offsets for topics A through Y will expire. > > > > How much do we care about this case? > > > > -James > > > > > On Jan 23, 2018, at 11:44 PM, Jeff Widman <j...@jeffwidman.com> wrote: > > > > > > Bumping this as I'd like to see it land... > > > > > > It's one of the "features" that tends to catch Kafka n00bs unawares > and > > > typically results in message skippage/loss, vs the proposed solution > is > > > much more intuitive behavior. > > > > > > Plus it's more wire efficient because consumers no longer need to > commit > > > offsets for partitions that have no new messages just to keep those > > offsets > > > alive. > > > > > > On Fri, Jan 12, 2018 at 10:21 AM, Vahid S Hashemian < > > > vahidhashem...@us.ibm.com> wrote: > > > > > >> There has been no further discussion on this KIP for about two > months. > > >> So I thought I'd provide the scoop hoping it would spark additional > > >> feedback and move the KIP forward. > > >> > > >> The KIP proposes a method to preserve group offsets as long as the > > group > > >> is not in Empty state (even when offsets are committed very rarely), > > and > > >> start the offset expiration of the group as soon as the group becomes > > >> Empty. > > >> It suggests dropping the `retention_time` field from the > `OffsetCommit` > > >> request and, instead, enforcing it via the broker config > > >> `offsets.retention.minutes` for all groups. In other words, all > groups > > >> will have the same retention time. > > >> The KIP presumes that this global retention config would suffice > common > > >> use cases and does not lead to, e.g., unmanageable offset cache size > > (for > > >> groups that don't need to stay around that long). It suggests opening > > >> another KIP if this global retention setting proves to be problematic > > in > > >> the future. It was suggested earlier in the discussion thread that > the > > KIP > > >> should propose a per-group retention config to circumvent this risk. > > >> > > >> I look forward to hearing your thoughts. Thanks! > > >> > > >> --Vahid > > >> > > >> > > >> > > >> > > >> From: "Vahid S Hashemian" <vahidhashem...@us.ibm.com> > > >> To: dev <dev@kafka.apache.org> > > >> Date: 10/18/2017 04:45 PM > > >> Subject: [DISCUSS] KIP-211: Revise Expiration Semantics of > > Consumer > > >> Group Offsets > > >> > > >> > > >> > > >> Hi all, > > >> > > >> I created a KIP to address the group offset expiration issue reported > > in > > >> KAFKA-4682: > > >> https://urldefense.proofpoint.com/v2/url?u=https-3A__cwiki. > > >> apache.org_confluence_display_KAFKA_KIP-2D211-253A-2BRevise- > > >> 2BExpiration-2BSemantics-2Bof-2BConsumer-2BGroup-2BOffsets& > > >> d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=Q_itwloTQj3_xUKl7Nzswo6KE4Nj- > > >> kjJc7uSVcviKUc&m=JkzH_2jfSMhCUPMk3rUasrjDAId6xbAEmX7_shSYdU4&s= > > >> UBu7D2Obulg0fterYxL5m8xrDWkF_O2kGlygTCWsfFc&e= > > >> > > >> > > >> Your feedback is welcome! > > >> > > >> Thanks. > > >> --Vahid > > >> > > >> > > >> > > >> > > >> > > >> > > > > > > > > > -- > > > > > > *Jeff Widman* > > > jeffwidman.com < > > https://urldefense.proofpoint.com/v2/url?u=http-3A__www. > > jeffwidman.com_&d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=Q_ > > itwloTQj3_xUKl7Nzswo6KE4Nj-kjJc7uSVcviKUc&m=TpGQEOVxEJ- > > > 3Y4rXTBP5CW7iUmO3PYKt_WeEDomWkAs&s=WXmFA_R-gtKbl9KgDfhHB4flnaBUB5C0ypRy4n > > 9xkoI&e= > > > | 740-WIDMAN-J (943-6265) > > > <>< > > > > > > > > > > > > > > > > >
