kafka-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From John Roesler <j...@confluent.io>
Subject Re: [DISCUSS] KIP-213: Second follow-up on Foreign Key Joins
Date Wed, 10 Jul 2019 15:58:48 GMT
Hey Matthias and Adam,

I've been mulling over your recent conversation. I'll share my two cents...

-----
First (because it's always best to get the semantics clear before the details):
Extra tombstones do NOT harm the semantics of an inner join. The
difference between a left join and an inner join result would be:
LEFT: (k -> (left-val, null))
INNER: (k -> null)

So, the tombstone result doesn't somehow make the inner join result
"left-join-like". Instead we're saying, "delete the join result for k
from the post-join view". It may be unnecessary to "delete" a
non-existant record from a view, but it's never incorrect. The issue
we're discussing is an operational one (whether the unnecessary
tombstones are a performance problem), not a semantic one.

Second, just to be clear, we have no obligation under current Streams
semantics to eliminate all unnecessary updates. We make no promise
whatsoever to "emit on change", so anyone consuming the output of a
Streams table operation should _expect_ to see unnecessary (aka
"duplicate") updates. We _can_ drop unnecessary updates (and do, when
the opportunity presents itself) to save on performance, though.

-----
Since it's clear that we're _only_ discussing a performance
optimization, we should try to characterize the actual performance
problem so that we can weigh the solution against the benefit it
yields.

Unless I'm mistaken, the exact problem we're discussing with the most
recent version of the proposal is that we'll emit an unnecessary
tombstone for inner joins or an unnecessary (left,null) result for
left joins specifically when a record transitions from one *broken* FK
reference to another *broken* FK reference. Any other transition
(valid FK -> valid FK, valid FK -> broken FK, or broken FK -> valid
FK) would yield the correct result (a necessary join result or a
necessary tombstone).

How likely are these transitions to occur in practice? We should
consider both the likelihood of FK changes at all and the likelihood
of broken FK references.

Probability of FK change:
It seems generally rare for records to change FK references at all.
How often to employees change departments, or comments change
articles, or people change addresses, etc, etc.? If someone has
anecdotal experience with a domain involving multiple, rapid FK
updates, please share them, but in all the domains I have experience
with, FK reference changes are _much_ less common than data updates. I
do happen to know that FK updates are moderately common in
Bazaarvoice's data model, but that they're orders of magnitude less
frequent than other kinds of updates.

Probability of broken reference:
Tables represented by Kafka topics (obviously) can't enforce
referential integrity, but (spitballing here) upstream databases can
and probably do. How often are we dealing with employees who work for
a manager who isn't in the HR database? Or a comment on a non-existent
article? Or someone who lives at a non-existent address? Again, this
is anecdotal, but that's the whole point of a likelihood estimation.
Please share your experiences. Again, I happen to know that at
Bazaarvoice, the incoming data actually doesn't guarantee referential
integrity, but that broken references are still relatively rare (and
actually, as it happens, it's not possible to change the FK without
first repairing a broken reference, so the broken -> broken transition
would be exceedingly rare).

If we can assume that the upstream data _mostly_ maintains referential
integrity, then we can make a further assumption that when we observe
a broken reference, it is probably transient. I.e., we can expect the
reference to be repaired "pretty soon". If this is the case, then the
probability we get a FK change from one broken reference to another
broken reference before the first reference gets repaired should be
quite rare for "normal" data sets.

Caveat: https://www.xkcd.com/2167/

-----

So, under the assumption that these unnecessary tombstones are rare,
and with the understanding that they're semantically ok to emit, it
really doesn't seem worthwhile to take on all the extra complexity
proposed earlier in this conversation.

A much simpler solution, if someone is really struggling with extra
tombstones, would be just be to slap an LRU cache on the result table
and drop unnecessary tombstones that way.

Again, just my personal thoughts, FWIW...
-John

On Tue, Jul 9, 2019 at 11:25 PM Adam Bellemare <adam.bellemare@gmail.com> wrote:
>
> I know what I posted was a bit of a wall of text, but three follow up
> thoughts to this:
>
> 1) Is it possible to enforce exactly-once for a portion of the topology? I
> was trying to think about how to process my proposal with at-least-once
> processing (or at-most-once processing) and I came up empty-handed.
>
> 2) A very deft solution is to also just support left-joins but not
> inner-joins. Practically speaking, either INNER or LEFT join as it
> currently is would support all of my use-cases.
>
> 3) Accept that there may be some null tombstones (though this makes me want
> to just go with LEFT only instead of LEFT and PSEUDO-INNER).
>
> In my experience (obviously empirical) it seems that many people just want
> the ability to join on foreign keys for the sake of handling all the
> relational data in their event streams and extra tombstones don't matter at
> all. This has been my own experience from our usage of our internal
> implementation at my company, and that of many others who have reached out
> to me.
>
> What would help most at this point is if someone can come up with a
> scenario where sending unnecessary tombstones actually poses a downstream
> problem beyond that of confusing behaviour, as I cannot think of one
> myself.  With that being said, I am far more inclined to actually then
> support just option #2 above and only have LEFT joins, forgoing INNER
> completely since it would not be a true inner join.
>
> Adam
>
>
>
>
>
>
>
>
> On Thu, Jul 4, 2019 at 8:50 AM Adam Bellemare <adam.bellemare@gmail.com>
> wrote:
>
> > Hi Matthias
> >
> > A thought about a variation of S1 that may work - it has a few moving
> > parts, so I hope I explained it clearly enough.
> >
> > When we change keys on the LHS:
> > (k,a) -> (k,b)
> >     (k,a,hashOf(b),PROPAGATE_OLD_VALUE) goes to RHS-0
> >     (k,b,PROPAGATE_NEW_VALUE) goes to RHS-1
> >
> > A) When the (k,a,hashOf(b),PROPAGATE_OLD_VALUE) hits RHS-0, the following
> > occurs:
> >   1) Store the current (CombinedKey<FK,K>, Value=(Hash, ForeignValue)) in
> > a variable
> >   2) Delete the key from the store
> >   3) Publish the event from step A-1 downstream with an instruction:
> > (eventType = COMPARE_TO_OTHER) (or whatever)
> > *      (key, (hashOf(b),wasForeignValueNull, eventType))*
> >     //Don't need the old hashOf(b) as it is guaranteed to be out of date
> >     //We do need the hashOf(b) that came with the event to be passed
> > along. Will be used in resolution.
> >     //Don't need the actual value as we aren't joining or comparing the
> > values, just using it to determine nulls. Reduces payload size.
> >
> > B) When (k,b,PROPAGATE_NEW_VALUE) hits RHS-1, the following occurs:
> >   1) Store it in the prefix state store (as we currently do)
> >   2) Get the FK-value (as we currently do)
> >   3) Return the normal SubscriptionResponse payload (eventType = UPDATE)
> > (or whatever)
> > *     (key, (hashOf(b), foreignValue, eventType))*
> >
> >
> > C) The Resolver Table is keyed on (as per our example):
> > key = CombinedKey<k-hash(b)>, value =
> > NullValueResolution<wasForeignValueNull (set by RHS-0), foreignValue (set
> > by RHS-1)>
> >
> > Resolution Steps per event:
> >
> > When one of either the output events from A (eventType ==
> > COMPARE_TO_OTHER) or B (eventType == UPDATE) is received
> > 1) Check if this event matches the current hashOf(b). If not, discard it,
> > for it is stale and no longer matters.  Additionally, delete entry
> > CombinedKey<k-hash(b)> from the Resolver Table.
> >
> > 2) Lookup event in table on its CombinedKey:
> >   - If it's not in the table, create the  NullValueResolution value,
> > populate the field related to the eventType, and add it to the table.
> >   - If it already IS in the table, get the existing NullValueResolution
> > object and finish populating it:
> >
> > 3) If the NullValueResolution is fully populated, move on to the
> > resolution logic below.
> >
> > Format:
> > (wasForeignValueNull, foreignValue) -> Result
> > If:
> > ( false  , Null ) -> Send tombstone. Old value was not null, new one is,
> > send tombstone.
> > (  true  , Null ) -> Do nothing.  See * below for more details.
> > (  true  , NewValue ) -> Send the new result
> > (  true  , NewValue ) -> Send the new result
> >
> > * wasForeignValueNull may have been false at some very recent point, but
> > only just translated to true (race condition). In this case, the RHS table
> > was updated and the value was set to null due to a an RHS update of (a,
> > oldVal) -> (a, null). This event on its own will propagate a delete event
> > through to the resolver (of a different eventType), so we don't need to
> > handle this case from the LHS and doing nothing is OK.
> >
> > In the case that it's truly (true, Null), we also don't need to send a
> > tombstone because wasForeignKeyNull == true means that a tombstone was
> > previously sent.
> >
> > 4) Check the hashOf(b) one last time before sending the resolved message
> > out. If the hash is old, discard it.
> >
> > 5) Delete the row from the Resolver Table.
> >
> >
> > Takeaways:
> > 1) I believe this is only necessary for INNER joins when we transition
> > from (k, non-Null-val) -> (k, non-Null-new-val)
> > 2) We can maintain state until we get both events back from RHS-0 and
> > RHS-1, at which point we delete it and clean up.
> >   NOTE: I think this still works with at-least-once processing, but I am
> > not 100% on this. The concern is that we receive events from RHS-0, RHS-1
> > (causing a row deletion), then RHS-1 again from at-least once (creating an
> > entry that never gets deleted).
> > 3) My naive implementation requires SubscriptionResponseWrapper event
> > types.
> > 4) This shouldn't affect performance much, as it should only be
> > maxOf(RHS-0, RHS-1) event propagation and processing.
> >
> >
> > Adam
> >
> > On Wed, Jul 3, 2019 at 2:18 PM Matthias J. Sax <matthias@confluent.io>
> > wrote:
> >
> >> It was KIP-77:
> >>
> >> https://cwiki.apache.org/confluence/display/KAFKA/KIP-77%3A+Improve+Kafka+Streams+Join+Semantics
> >>
> >> It did all kind of improvements, including removing unnecessary
> >> tombstones.
> >>
> >> About S-1: I agree that it would be complex and impact performance. I
> >> did not think about all details yet, but want to throw out the idea first.
> >>
> >> > S-2 could probably be simplified to "for a given key, was the previous
> >> > propagated result a null/tombstone or not?"
> >>
> >> Yes, that's the idea.
> >>
> >> > However,
> >> > the table will grow indefinitely large as we can never remove keys from
> >> it.
> >>
> >> That a good point. I missed that... Seems that this rules out S-2
> >> because if we cannot provide _strict_ guarantees, it does not seem to be
> >> worth to even try?
> >>
> >>
> >> About the impact of unnecessary tombstones: There is a performance
> >> impact, as we increase the output data rate, potentially broker load if
> >> the result is written into a topic, and also increase downstream
> >> processing load. It's hard to judge, how much the overhead will be, as
> >> it will depend on the selectivity of the join. But it might be
> >> significant?
> >>
> >> Also, users have certain expectations on the result and it's unintuitive
> >> (even if semantically correct) to send those tombstones. From
> >> experience, we often have a hard time to explain semantics to people and
> >> I was hoping we could avoid introducing unintuitive behavior.
> >>
> >>
> >> Would be good to get input from others and how they judge the impact. I
> >> think it might be still worth to explore how complex S-1 would be. If we
> >> think it's too complex it might be a good argument to just accept the
> >> unnecessary tombstones?
> >>
> >>
> >> -Matthias
> >>
> >>
> >> On 7/3/19 8:03 AM, Adam Bellemare wrote:
> >> > Hi Matthias
> >> >
> >> > Do you happen to recall what the impact was of having unnecessary
> >> > tombstones? I am wondering if the negative impact is still relevant
> >> today,
> >> > and if so, if you can recall the PRs or KIPs related to it.
> >> >
> >> > That being said, I think that S-1 is too complex in terms of
> >> > synchronization. It seems to me that the processor would need to block
> >> > while it waits for the unsubscribe to propagate and return, which would
> >> > cause throughput to drop significantly. Alternately, we would need to
> >> > maintain state anyways about which events were sent and which responses
> >> > returned, while being sure to respect the offset order in which they're
> >> > emitted. I think this would only reduce blocking slightly while
> >> increasing
> >> > complexity. If I am wrong in understanding this, please let me know
> >> where
> >> > my thinking is erroneous.
> >> >
> >> > S-2 could probably be simplified to "for a given key, was the previous
> >> > propagated result a null/tombstone or not?"
> >> > It would act very similarly to the hash value mechanism, where we
> >> discard
> >> > any events that are not of the correct hash. In this case, we simply
> >> store
> >> > (key, wasLastOutputATombstone) right before the event is output
> >> downstream
> >> > of the Join + Resolver. This ignores all the complexities of which
> >> event is
> >> > propagating over which wire and simply squelches any extra tombstones
> >> from
> >> > being sent.
> >> >
> >> > For storage, we need to use the full primary key and a boolean. However,
> >> > the table will grow indefinitely large as we can never remove keys from
> >> it.
> >> > If we delete key=k from the table and propagate a tombstone, but later
> >> (say
> >> > 3 weeks, 3 months, etc) we publish (k, baz), but baz does not exist on
> >> the
> >> > RHS, we will end up publishing an extra tombstone because we have no
> >> idea
> >> > what the previously sent record was for k. For this reason I think it's
> >> > worth asking if we really can maintain state, and if it's even necessary
> >> > (again, a full understanding of the impact of extra tombstones may help
> >> us
> >> > figure out a better solution).
> >> >
> >> > As it stands, I don't think either of these will work well. That being
> >> > said, I myself do not have any better ideas at the moment, but I would
> >> > still like to better understand circumstances where it has a negative
> >> > impact downstream as that may provide some insights.
> >> >
> >> >
> >> > Thanks
> >> >
> >> > Adam
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > On Tue, Jul 2, 2019 at 11:18 PM Matthias J. Sax <matthias@confluent.io>
> >> > wrote:
> >> >
> >> >> Thanks for the example. I was thinking about the problem a little bit,
> >> >> and I believe we should look at it in some more details.
> >> >>
> >> >> Basically, there are 3 cases:
> >> >>
> >> >> a) insert new record LHS
> >> >> b) delete record LHS
> >> >> c) update exiting record LHS
> >> >>
> >> >> For those cases we want different things to happen:
> >> >>
> >> >> a-1) sent subscribe message to RHS
> >> >> a-2) RHS lookup and send result back if there is one
> >> >> a-3) emit result on LHS if any is returned
> >> >>
> >> >> b-1) delete subscription from RHS
> >> >> b-2) if there was a previous result (can easily be decided by looking
> >> up
> >> >> RHS table for an existing key), send tombstone back
> >> >> b-3) emit tombstone on LHS if any is returned
> >> >>
> >> >> c-1) delete old subscription from RHS
> >> >> c-2) send new subscription to RHS
> >> >> c-3) if there was no previous result and there is no new result emit
> >> >> nothing
> >> >> c-4) if there was a previous result and there is no new result emit a
> >> >> tombstone LHS
> >> >> c-5) if there is a new result (old result may or may not exist), emit
> >> >> only new result LHS (don't emit a tombstone)
> >> >>
> >> >>
> >> >> Case (a) and (b) are simple and could be implemented with a "fire and
> >> >> forget" strategy. The LHS just "blindly" updates the subscription, the
> >> >> RHS can process the subscription with local knowledge and may send
> >> >> something back. If something is sent back, the LHS blindly emits it.
> >> >>
> >> >> We know that for both cases, we never miss anything and we never emit
> >> >> anything unnecessary.
> >> >>
> >> >> However, using this pattern for (c), we don't get our expected result:
> >> >>
> >> >> Issues: LHS sends both subscription updates in parallel. It does not
> >> >> know if zero, one, or two result records will be produced by RHS. If
> >> RHS
> >> >> produces two results, their order is not known (however, LHS would need
> >> >> to emit them in the right order; also forcing RHS to always send a
> >> >> result back is not a sufficient solution). If only one result comes
> >> >> back, it's unclear if a second result may arrive later and thus the
> >> >> result may need to be buffered... Overall, local knowledge does not
> >> seem
> >> >> to be sufficient to tackle the case.
> >> >>
> >> >> The current proposal tries to address the issue with encoding
> >> additional
> >> >> information, to tell the RHS to send a tombstone back or not. But it
> >> >> does not seem to be perfect, and it might result in unnecessary
> >> >> tombstones as it still uses local knowledge only and thus misses some
> >> >> information.
> >> >>
> >> >> I think the main problem is, that the knowledge about the a potential
> >> >> previous result and a potential new result is sharded on the RHS.
> >> Hence,
> >> >> the "unsubscribe" does not know if it needs to send a tombstone back
> >> for
> >> >> the case that there was an old result but there is no new result.
> >> >> Similarly, the "new subscribe" cannot know if it needs to send a
> >> >> tombstone or not (as it does not know if there was a previous result)
> >> if
> >> >> it does not match.
> >> >>
> >> >> To really solve the issue, I see two possible solutions (both are not
> >> >> great, but I wanted to discuss them anyway):
> >> >>
> >> >> S-1: First unsubscribe, and send new subscription after result comes
> >> >> back. For this case, the RHS must always send something back to the LHS
> >> >> on unsubscribe. The answer if "previous result exists/not-exist" can be
> >> >> added to the new-subscription and hence RHS can either return nothing,
> >> a
> >> >> tombstone, or a new result. The LHS can blindly emit whatever RHS
> >> >> returns. This would also cover (a) and (b) cases. However, the overall
> >> >> time to emit the join result is doubled for the (common) update case...
> >> >> (we need two consecutive round-trips to the RHS).
> >> >>
> >> >> S-2: Remember/store if a previous result exists on LHS: for this case,
> >> >> (a) is handled straightforward, (b) is handled by telling RHS to send
> >> >> tombstone if previous result exits, and (c) can send both request in
> >> >> parallel letting the unsubscribe never return anything, and subscribe
> >> is
> >> >> handled as in (b). However, we need a second store on the LHS to
> >> >> remember if there was a previous result. (Also not sure how
> >> >> interleaving/inflight computation might affect the algorithm...)
> >> >>
> >> >> I think, sending unnecessary tombstones is quite bad (in very old
> >> >> releases we had a similar issue and fixed it). However, I am also not
> >> >> 100% sure if the solutions I came up with are good enough to justify
> >> >> them. (Personally, I slightly tend to prefer S-2 because I think that
> >> >> the additional store is less of an issue than the increase processing
> >> >> time).
> >> >>
> >> >> Would love to hear your thoughts.
> >> >>
> >> >>
> >> >> -Matthias
> >> >>
> >> >>
> >> >> On 6/28/19 6:19 AM, Adam Bellemare wrote:
> >> >>> Hi Matthias
> >> >>>
> >> >>> Yes, thanks for the questions - I know it's hard to keep up with all
> >> of
> >> >> the
> >> >>> various KIPs and everything.
> >> >>>
> >> >>> The instructions are not stored anywhere, but are simply a way of
> >> letting
> >> >>> the RHS know how to handle the subscription and reply accordingly.
> >> >>>
> >> >>> The only case where we send an unnecessary tombstone is (that I can
> >> >>> tell...) when we do the following:
> >> >>> RHS:
> >> >>> (1, bar)
> >> >>>
> >> >>> LHS
> >> >>> (K,1)  -> Results in (K, 1, bar) being output
> >> >>> (K,1) -> (K,2) ->  Results in (K, null) being output for INNER (no
> >> >> matching
> >> >>> element on LHS)
> >> >>> (K,2) -> (K,3) ->  Results in (K, null) being output for INNER
> >> (because
> >> >> we
> >> >>> don't maintain state to know we already output the tombstone on the
> >> >>> previous transition).
> >> >>> (K,2) -> (K,9000) ->  Results in (K, null)... etc.
> >> >>>
> >> >>> Byte versioning is going in today, then I hope to get back to
> >> addressing
> >> >> a
> >> >>> number of John's previous questions in the PR.
> >> >>>
> >> >>> Adam
> >> >>>
> >> >>>
> >> >>> On Thu, Jun 27, 2019 at 5:47 PM Matthias J. Sax <
> >> matthias@confluent.io>
> >> >>> wrote:
> >> >>>
> >> >>>> Thanks for bringing this issue to our attention. Great find @Joe!
> >> >>>>
> >> >>>> Adding the instruction field to the `subscription` sounds like a good
> >> >>>> solution. What I don't understand atm: for which case would we need
> >> to
> >> >>>> send unnecessary tombstone? I thought that the `instruction` field
> >> helps
> >> >>>> to avoid any unnecessary tombstone? Seems I a missing case?
> >> >>>>
> >> >>>> Also for my own understanding: the `instruction` is only part of the
> >> >>>> message? It is no necessary to store it in the RHS auxiliary store,
> >> >> right?
> >> >>>>
> >> >>>> About right/full-outer joins. Agreed. Getting left-joins would be
> >> >> awesome!
> >> >>>>
> >> >>>> About upgrading: Good call John! Adding a version byte for
> >> subscription
> >> >>>> and response is good forward thinking. I personally prefer version
> >> >>>> numbers, too, as they carry more information.
> >> >>>>
> >> >>>> Thanks for all the hard to everybody involved!
> >> >>>>
> >> >>>>
> >> >>>> -Matthias
> >> >>>>
> >> >>>> On 6/27/19 1:44 PM, John Roesler wrote:
> >> >>>>> Hi Adam,
> >> >>>>>
> >> >>>>> Hah! Yeah, I felt a headache coming on myself when I realized this
> >> >>>>> would be a concern.
> >> >>>>>
> >> >>>>> For what it's worth, I'd also lean toward versioning. It seems more
> >> >>>>> explicit and more likely to keep us all sane in the long run. Since
> >> we
> >> >>>>> don't _think_ our wire protocol will be subject to a lot of
> >> revisions,
> >> >>>>> we can just use one byte. The worst case is that we run out of
> >> numbers
> >> >>>>> and reserve the last one to mean, "consult another field for the
> >> >>>>> actual version number". It seems like a single byte on each message
> >> >>>>> isn't too much to pay.
> >> >>>>>
> >> >>>>> Since you point it out, we might as well put a version number on the
> >> >>>>> SubscriptionResponseWrapper as well. It may not be needed, but if we
> >> >>>>> ever need it, even just once, we'll be glad we have it.
> >> >>>>>
> >> >>>>> Regarding the instructions field, we can also serialize the enum
> >> very
> >> >>>>> compactly as a single byte (which is the same size a boolean takes
> >> >>>>> anyway), so it seems like an Enum in Java-land and a byte on the
> >> wire
> >> >>>>> is a good choice.
> >> >>>>>
> >> >>>>> Agreed on the right and full outer joins, it doesn't seem necessary
> >> >>>>> right now, although I am happy to see the left join "join" the
> >> party,
> >> >>>>> since as you said, we were so close to it anyway. Can you also add
> >> it
> >> >>>>> to the KIP?
> >> >>>>>
> >> >>>>> Thanks as always for your awesome efforts on this,
> >> >>>>> -John
> >> >>>>>
> >> >>>>> On Thu, Jun 27, 2019 at 3:04 PM Adam Bellemare <
> >> >> adam.bellemare@gmail.com>
> >> >>>> wrote:
> >> >>>>>>
> >> >>>>>> You're stretching my brain, John!
> >> >>>>>>
> >> >>>>>> I prefer STRATEGY 1 because it solves the problem in a simple way,
> >> and
> >> >>>>>> allows us to deprecate support for older message types as we go
> >> (ie,
> >> >> we
> >> >>>>>> only support the previous 3 versions, so V5,V4,V3, but not v2 or
> >> V1).
> >> >>>>>>
> >> >>>>>> STRATEGY 2 is akin to Avro schemas between two microservices -
> >> there
> >> >> are
> >> >>>>>> indeed cases where a breaking change must be made, and forward
> >> >>>>>> compatibility will provide us with no out other than requiring a
> >> full
> >> >>>> stop
> >> >>>>>> and full upgrade for all nodes, shifting us back towards STRATEGY
> >> 1.
> >> >>>>>>
> >> >>>>>> My preference is STRATEGY 1 with instructions as an ENUM, and we
> >> can
> >> >>>>>> certainly include a version. Would it make sense to include a
> >> version
> >> >>>>>> number in  SubscriptionResponseWrapper as well? Currently we don't
> >> >> have
> >> >>>> any
> >> >>>>>> instructions in there, as I removed the boolean, but it is
> >> certainly
> >> >>>>>> plausible that it could happen in the future. I don't *think* we'll
> >> >> need
> >> >>>>>> it, but I also didn't think we'd need it for SubscriptionWrapper
> >> and
> >> >>>> here
> >> >>>>>> we are.
> >> >>>>>>
> >> >>>>>> Thanks for the thoughts, and the info on the right-key. That was
> >> >>>>>> enlightening, though I can't think of a use-case for it *at this
> >> point
> >> >>>> in
> >> >>>>>> time*. :)
> >> >>>>>>
> >> >>>>>> Adam
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>
> >> >>>>>> On Thu, Jun 27, 2019 at 12:29 PM John Roesler <john@confluent.io>
> >> >>>> wrote:
> >> >>>>>>
> >> >>>>>>> I think I agree with you, right joins (and therefore full outer
> >> >> joins)
> >> >>>>>>> don't make sense here, because the result is a keyed table, where
> >> the
> >> >>>>>>> key is the PK of the left-hand side. So, when you have a
> >> >>>>>>> right-hand-side record with no incoming FK references, you would
> >> want
> >> >>>>>>> to produce a join result like `nullKey: (null, rhsValue)`, but we
> >> >>>>>>> don't currently allow null keys in Streams. It actually is
> >> possible
> >> >> to
> >> >>>>>>> define them, and therefore to add right- and full-outer
> >> foreign-key
> >> >>>>>>> joins later, but it's non-trivial in a streaming context with
> >> >>>>>>> continuously updated results. (See the PS if you're curious what
> >> I'm
> >> >>>>>>> thinking). You're correct, right- and full-outer joins are
> >> trivial on
> >> >>>>>>> our current 1:1 table joins because they are equi-joins.
> >> >>>>>>>
> >> >>>>>>> Regarding the transition, it sounds like what you're proposing is
> >> >> that
> >> >>>>>>> we would say, "adding a foreign-key join to your topology
> >> requires a
> >> >>>>>>> full application reset (or a new application id)". This is also an
> >> >>>>>>> acceptable constraint to place on the feature, but not strictly
> >> >>>>>>> necessary. Since 2.3, it's now possible to give all the state in
> >> your
> >> >>>>>>> application stable names. This means that it's no longer true that
> >> >>>>>>> adding a node to your topology graph would break its structure,
> >> and
> >> >> it
> >> >>>>>>> does become possible to add new operators and simply restart the
> >> app.
> >> >>>>>>> Revisiting my prior thought, though, I think the problem is not
> >> >>>>>>> specific to your feature. For example, adding a new grouped
> >> >>>>>>> aggregation would produce a new repartition topic, but the
> >> >> repartition
> >> >>>>>>> topic partitions might get assigned to old nodes in the middle of
> >> a
> >> >>>>>>> rolling bounce, and they would need to just ignore them. This
> >> >>>>>>> requirement is the same for the repartition topics in the FK
> >> join, so
> >> >>>>>>> it's orthogonal to your design.
> >> >>>>>>>
> >> >>>>>>> Back to the first concern, though, I'm not sure I followed the
> >> >>>>>>> explanation. As a thought experiment, let's imagine that Joe
> >> hadn't
> >> >>>>>>> taken the time to experiment with your feature branch. We wouldn't
> >> >>>>>>> have noticed the problem until the feature was already released in
> >> >>>>>>> 2.4. So the wire protocol on that PK->FK subscription topic would
> >> >> have
> >> >>>>>>> been V1: "FK,PK,HASH,BOOLEAN". Then, Joe would have let us know
> >> the
> >> >>>>>>> problem once they picked up the feature, so we would want to
> >> >> implement
> >> >>>>>>> your proposed fix and change the wire protocol to V2:
> >> >>>>>>> "FK,PK,HASH,INSTRUCTIONS" in 2.5. Upon rolling out the update, we
> >> >>>>>>> would see both 2.4 nodes encountering V2 messages and 2.5 nodes
> >> >>>>>>> encountering V1 messages. How can they both detect that they are
> >> >>>>>>> attempting to process a newer or older protocol? If they can
> >> detect
> >> >>>>>>> it, then what should they do?
> >> >>>>>>>
> >> >>>>>>> From experience, there are two basic solutions to this problem:
> >> >>>>>>>
> >> >>>>>>> STRATEGY1. Add a protocol version to the message (could be a
> >> number
> >> >> at
> >> >>>>>>> the start of the message payload, or it could be a number in the
> >> >>>>>>> message headers, not sure if it matters much. Payload is probably
> >> >> more
> >> >>>>>>> compact, since the header would need a name.) In this case, the
> >> 2.4
> >> >>>>>>> worker would know that it's max protocol version is V1, and when
> >> it
> >> >>>>>>> sees the V2 message, it knows that it can't handle it properly.
> >> >> Rather
> >> >>>>>>> than doing something wrong, it would just not do anything. This
> >> means
> >> >>>>>>> it would stop the task, if not shut down the whole instance. On
> >> the
> >> >>>>>>> other hand, a 2.5 worker would have some defined logic for how to
> >> >>>>>>> handle all versions (V1 and V2), so once the upgrade is complete,
> >> all
> >> >>>>>>> messages can be processed.
> >> >>>>>>>
> >> >>>>>>> STRATEGY2. Make the schema forward-compatible. Basically, we
> >> ensure
> >> >>>>>>> that new fields can only be appended to the message schema, and
> >> that
> >> >>>>>>> older workers using only a prefix of the full message would still
> >> >>>>>>> behave correctly. Using the example above, we'd instead evolve the
> >> >>>>>>> schema to V2': "FK,PK,HASH,BOOLEAN,INSTRUCTIONS", and continue to
> >> set
> >> >>>>>>> the boolean field to true for the "new" foreign key. Then, 2.4
> >> >> workers
> >> >>>>>>> encountering the a "new FK" message would just see the prefix of
> >> the
> >> >>>>>>> payload that makes sense to them, and they would still continue
> >> >>>>>>> processing the messages as they always have. Only after the 2.5
> >> code
> >> >>>>>>> is fully rolled out to the cluster would we be sure to see the
> >> >> desired
> >> >>>>>>> behavior. Note: in the reverse case, a 2.5 worker knows how to
> >> fully
> >> >>>>>>> parse the new message format, even if it plans to ignore the
> >> BOOLEAN
> >> >>>>>>> field.
> >> >>>>>>>
> >> >>>>>>> There are some tradeoffs between these strategies: STRATEGY1
> >> ensures
> >> >>>>>>> that all messages are only handled by workers that can properly
> >> >> handle
> >> >>>>>>> them, although it results in processing stalls while there are
> >> still
> >> >>>>>>> old nodes in the cluster. STRATEGY2 ensures that all messages can
> >> be
> >> >>>>>>> processed by all nodes, so there are no stalls, but we can never
> >> >>>>>>> remove fields from the message, so if there are a lot of
> >> revisions in
> >> >>>>>>> the future, the payloads will become bloated. Also, it's not clear
> >> >>>>>>> that you can actually pull off STRATEGY2 in all cases. If there's
> >> >> some
> >> >>>>>>> new kind of message you want to send that has no way to be
> >> correctly
> >> >>>>>>> processed at all under the 2.4 code paths, the prefix thing simply
> >> >>>>>>> doesn't work. Etc.
> >> >>>>>>>
> >> >>>>>>> Also, note that you can modify the above strategies by instead
> >> >>>>>>> designing the message fields for extensibility. E.g., if you make
> >> the
> >> >>>>>>> instructions field an enum, then you can make sure that the
> >> default
> >> >>>>>>> case is handled sensibly (probably similarly to STRATEGY1, just
> >> choke
> >> >>>>>>> on unknown instructions) and that you never remove an instruction
> >> >> type
> >> >>>>>>> from the enum in future versions.
> >> >>>>>>>
> >> >>>>>>> Does this make sense?
> >> >>>>>>> -John
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>> PS:
> >> >>>>>>> We can define null keys for streaming tables, but it's tricky.
> >> >>>>>>>
> >> >>>>>>> Specifically, you'd want to define some concept of null keys that
> >> >>>>>>> allows all null keys to be unique, but _also_ to have a fixed
> >> >>>>>>> identity, so that a particular null-key can be updated later. One
> >> >>>>>>> example could be to union the existing keyspace with a new
> >> >>>>>>> null-keyspace, where normal keys are like "key" and null-keys are
> >> >> like
> >> >>>>>>> "null(identity)". Then given a query like
> >> >>>>>>> "KTable<String,Integer>.rightJoin(KTable<Integer,Boolean>)", and
> >> >>>>>>> inputs like:
> >> >>>>>>> LHS:
> >> >>>>>>> "a": 1
> >> >>>>>>> "b": 2
> >> >>>>>>>
> >> >>>>>>> RHS:
> >> >>>>>>> 1: true
> >> >>>>>>> 3: false
> >> >>>>>>>
> >> >>>>>>> a full outer join would produce:
> >> >>>>>>> "a": (1, true)
> >> >>>>>>> "b": (2, null)
> >> >>>>>>> null(3): (null, false)
> >> >>>>>>>
> >> >>>>>>> which can be correctly updated later if we get an update on the
> >> LHS:
> >> >>>>>>> PUT("c": 3)
> >> >>>>>>>
> >> >>>>>>> We'd emit for the results:
> >> >>>>>>> DELETE(null(e))
> >> >>>>>>> EMIT("c": (3, false))
> >> >>>>>>>
> >> >>>>>>> Resulting in the correct result table of:
> >> >>>>>>> "a": (1, true)
> >> >>>>>>> "b": (2, null)
> >> >>>>>>> "c": (3, false)
> >> >>>>>>>
> >> >>>>>>> As mentioned, this is tricky, and I would avoid it until we have
> >> >>>>>>> evidence that it's actually useful to cover this part of the
> >> design
> >> >>>>>>> space. Certainly, it would be a separate KIP if it came to that.
> >> >>>>>>>
> >> >>>>>>> On Wed, Jun 26, 2019 at 8:57 PM Adam Bellemare <
> >> >>>> adam.bellemare@gmail.com>
> >> >>>>>>> wrote:
> >> >>>>>>>>
> >> >>>>>>>> Hi John
> >> >>>>>>>>
> >> >>>>>>>> Good thinking with regards to upgrade path between versions
> >> >> regarding
> >> >>>>>>>> over-the-wire instructions in SubscriptionWrapper. At this point
> >> in
> >> >>>> time
> >> >>>>>>> I
> >> >>>>>>>> can't think of any new wire message instructions, but I would
> >> >>>> appreciate
> >> >>>>>>> as
> >> >>>>>>>> many eyes on it as possible. I have just included the LEFT join
> >> in
> >> >> the
> >> >>>>>>> last
> >> >>>>>>>> commit (about 10 min ago) along with INNER join. I do not think
> >> that
> >> >>>>>>> RIGHT
> >> >>>>>>>> join and OUTER are possible given that there is no LHS key
> >> >> available,
> >> >>>> so
> >> >>>>>>>> LHSTable.outerJoinOnForeignKey(RHSTable) wouldn't even make
> >> sense.
> >> >>>> This
> >> >>>>>>> is
> >> >>>>>>>> in contrast to the current LHSTable.outerJoin(RHSTable), as they
> >> are
> >> >>>> both
> >> >>>>>>>> keyed on the same key. I have buffed up the Integration tests and
> >> >> have
> >> >>>>>>>> tried to make them more readable to ensure that we're covering
> >> all
> >> >> the
> >> >>>>>>>> scenarios. I think that if we can get more eyes on the workflow
> >> >>>> showing
> >> >>>>>>> the
> >> >>>>>>>> various LHS and RHS events and outputs then that may help us
> >> >> validate
> >> >>>>>>> that
> >> >>>>>>>> we have all the scenarios covered.
> >> >>>>>>>>
> >> >>>>>>>> With regards to the 2.3->2.4 scenario you described, I'm not
> >> >> entirely
> >> >>>>>>> sure
> >> >>>>>>>> I follow. If they want to add a FK-join, they will need to rework
> >> >>>> their
> >> >>>>>>>> code in the KStreams app and make a new release, since the
> >> >> underlying
> >> >>>>>>>> topology would be different and new internal topics would need
> >> to be
> >> >>>>>>>> created. In other words, I don't think a rolling upgrade where
> >> the
> >> >>>> user
> >> >>>>>>>> introduces a FK join would be possible since their topology would
> >> >>>>>>>> necessitate a full KStreams reset. Is this what you meant?
> >> >>>>>>>>
> >> >>>>>>>>
> >> >>>>>>>>
> >> >>>>>>>> On Wed, Jun 26, 2019 at 4:10 PM John Roesler <john@confluent.io>
> >> >>>> wrote:
> >> >>>>>>>>
> >> >>>>>>>>> Thanks, Adam!
> >> >>>>>>>>>
> >> >>>>>>>>> One unrelated thought that has just now occurred to me is that
> >> >>>> (unlike
> >> >>>>>>>>> the equi-joins we currently have), this join logic is
> >> potentially
> >> >>>>>>>>> spread over multiple Streams instances, which in general means
> >> that
> >> >>>>>>>>> the instances may be running different versions of Kafka
> >> Streams.
> >> >>>>>>>>>
> >> >>>>>>>>> This means that if we discover a bug that requires us to again
> >> >> change
> >> >>>>>>>>> the wire message (as you did in this proposal update), we need
> >> to
> >> >>>>>>>>> consider what should happen if the PK instance is newer than
> >> the FK
> >> >>>>>>>>> instance, or vice-versa, during a rolling upgrade. We should
> >> think
> >> >>>>>>>>> ahead to this condition and make sure the logic is forward
> >> >>>> compatible.
> >> >>>>>>>>>
> >> >>>>>>>>> Related: what about the initial case, when we release this
> >> feature
> >> >>>>>>>>> (let's say in 2.4)? What will happen if I decide to adopt 2.4
> >> and
> >> >> add
> >> >>>>>>>>> a FK join together in one upgrade. Thus, the 2.4 member of the
> >> >>>> cluster
> >> >>>>>>>>> is producing the SubscriptionWrapper messages, and some 2.3
> >> members
> >> >>>>>>>>> get the subscription topic assigned to them, but they have no
> >> idea
> >> >>>>>>>>> what to do with it? I'm not sure this is a problem; hopefully
> >> they
> >> >>>>>>>>> just do nothing. If it is a problem, it would be fine to say you
> >> >> have
> >> >>>>>>>>> to upgrade completely to 2.4 before deploying a FK join.
> >> >>>>>>>>>
> >> >>>>>>>>> Just want to make sure we anticipate these issues in case it
> >> >> affects
> >> >>>>>>>>> the design at all.
> >> >>>>>>>>>
> >> >>>>>>>>> Thanks,
> >> >>>>>>>>> -John
> >> >>>>>>>>>
> >> >>>>>>>>> On Wed, Jun 26, 2019 at 2:38 PM Adam Bellemare <
> >> >>>>>>> adam.bellemare@gmail.com>
> >> >>>>>>>>> wrote:
> >> >>>>>>>>>>
> >> >>>>>>>>>> Sigh... Forgot the link:
> >> >>>>>>>>>>
> >> >>>>>>>>>
> >> >>>>>>>
> >> >>>>
> >> >>
> >> https://cwiki.apache.org/confluence/pages/diffpagesbyversion.action?pageId=74684836&selectedPageVersions=78&selectedPageVersions=74
> >> >>>>>>>>>>
> >> >>>>>>>>>> I'll update it when I validate that there are no issues with
> >> >>>>>>> removing the
> >> >>>>>>>>>> SubscriptionResponseWrapper boolean.
> >> >>>>>>>>>>
> >> >>>>>>>>>> On Wed, Jun 26, 2019 at 3:37 PM Adam Bellemare <
> >> >>>>>>> adam.bellemare@gmail.com
> >> >>>>>>>>>>
> >> >>>>>>>>>> wrote:
> >> >>>>>>>>>>
> >> >>>>>>>>>>>> Maybe just call it as (k, leftval, null) or (k, null,
> >> rightval)?
> >> >>>>>>>>>>> Done.
> >> >>>>>>>>>>>
> >> >>>>>>>>>>>> if you update the KIP, you might want to send a new "diff
> >> link"
> >> >>>>>>> to
> >> >>>>>>>>> this
> >> >>>>>>>>>>> thread
> >> >>>>>>>>>>> Here it is:
> >> >>>>>>>>>>>
> >> >>>>>>>>>>>> Looking closely at the proposal, can you explain more about
> >> the
> >> >>>>>>>>>>> propagateIfNull field in SubscriptionResponseWrapper? It sort
> >> of
> >> >>>>>>> looks
> >> >>>>>>>>> like
> >> >>>>>>>>>>> it's always going to be equal to (RHS-result != null).
> >> >>>>>>>>>>> I believe you are correct, and I missed the forest for the
> >> trees.
> >> >>>>>>> They
> >> >>>>>>>>> are
> >> >>>>>>>>>>> effectively the same thing, and I can simply remove the flag.
> >> I
> >> >>>>>>> will
> >> >>>>>>>>> code
> >> >>>>>>>>>>> it up and try it out locally just to be sure.
> >> >>>>>>>>>>>
> >> >>>>>>>>>>> Thanks again for your help, it is greatly appreciated!
> >> >>>>>>>>>>>
> >> >>>>>>>>>>> On Wed, Jun 26, 2019 at 2:54 PM John Roesler <
> >> john@confluent.io>
> >> >>>>>>>>> wrote:
> >> >>>>>>>>>>>
> >> >>>>>>>>>>>> I think the "scenario trace" is very nice, but has one point
> >> >> that
> >> >>>>>>> I
> >> >>>>>>>>>>>> found confusing:
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> You indicate a retraction in the join output as (k,null) and
> >> a
> >> >>>>>>> join
> >> >>>>>>>>>>>> result as (k, leftval, rightval), but confusingly, you also
> >> >> write
> >> >>>>>>> a
> >> >>>>>>>>>>>> join result as (k, JoinResult) when one side is null. Maybe
> >> just
> >> >>>>>>> call
> >> >>>>>>>>>>>> it as (k, leftval, null) or (k, null, rightval)? That way the
> >> >>>>>>> readers
> >> >>>>>>>>>>>> can more easily determine if the results meet their
> >> expectations
> >> >>>>>>> for
> >> >>>>>>>>>>>> each join type.
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> (procedural note: if you update the KIP, you might want to
> >> send
> >> >> a
> >> >>>>>>> new
> >> >>>>>>>>>>>> "diff link" to this thread, since the one I posted at the
> >> >>>>>>> beginning
> >> >>>>>>>>>>>> would not automatically show your latest changes)
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> I was initially concerned that the proposed algorithm would
> >> wind
> >> >>>>>>> up
> >> >>>>>>>>>>>> propagating something that looks like a left join (k,
> >> leftval,
> >> >>>>>>> null)
> >> >>>>>>>>>>>> under the case that Joe pointed out, but after reviewing your
> >> >>>>>>>>>>>> scenario, I see that it will emit a tombstone (k, null)
> >> instead.
> >> >>>>>>> This
> >> >>>>>>>>>>>> is appropriate, and unavoidable, since we have to retract the
> >> >> join
> >> >>>>>>>>>>>> result from the logical view (the join result is a logical
> >> >> Table).
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> Looking closely at the proposal, can you explain more about
> >> the
> >> >>>>>>>>>>>> propagateIfNull field in SubscriptionResponseWrapper?
> >> >>>>>>>>>>>> It sort of looks like it's always going to be equal to
> >> >>>>>>> (RHS-result !=
> >> >>>>>>>>>>>> null).
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> In other words, can we drop that field and just send back
> >> >>>>>>> RHS-result
> >> >>>>>>>>>>>> or null, and then handle it on the left-hand side like:
> >> >>>>>>>>>>>> if (rhsOriginalValueHash doesn't match) {
> >> >>>>>>>>>>>>     emit nothing, just drop the update
> >> >>>>>>>>>>>> } else if (joinType==inner && rhsValue == null) {
> >> >>>>>>>>>>>>     emit tombstone
> >> >>>>>>>>>>>> } else {
> >> >>>>>>>>>>>>     emit joiner(lhsValue, rhsValue)
> >> >>>>>>>>>>>> }
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> To your concern about emitting extra tombstones, personally,
> >> I
> >> >>>>>>> think
> >> >>>>>>>>>>>> it's fine. Clearly, we should try to avoid unnecessary
> >> >>>>>>> tombstones, but
> >> >>>>>>>>>>>> all things considered, it's not harmful to emit some
> >> unnecessary
> >> >>>>>>>>>>>> tombstones: their payload is small, and they are trivial to
> >> >> handle
> >> >>>>>>>>>>>> downstream. If users want to, they can materialize the join
> >> >>>>>>> result to
> >> >>>>>>>>>>>> suppress any extra tombstones, so there's a way out.
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> Thanks for the awesome idea. It's better than what I was
> >> >> thinking.
> >> >>>>>>>>>>>> -john
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>> On Wed, Jun 26, 2019 at 11:37 AM Adam Bellemare
> >> >>>>>>>>>>>> <adam.bellemare@gmail.com> wrote:
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>> Thanks John.
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>> I'm looking forward to any feedback on this. In the
> >> meantime I
> >> >>>>>>> will
> >> >>>>>>>>>>>> work on
> >> >>>>>>>>>>>>> the unit tests to ensure that we have well-defined and
> >> readable
> >> >>>>>>>>>>>> coverage.
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>> At the moment I cannot see a way around emitting (k,null)
> >> >>>>>>> whenever
> >> >>>>>>>>> we
> >> >>>>>>>>>>>> emit
> >> >>>>>>>>>>>>> an event that lacks a matching foreign key on the RHS,
> >> except
> >> >>>>>>> in the
> >> >>>>>>>>>>>>> (k,null) -> (k,fk) case.
> >> >>>>>>>>>>>>> If this LHS oldValue=null, we know we would have emitted a
> >> >>>>>>> deletion
> >> >>>>>>>>> and
> >> >>>>>>>>>>>> so
> >> >>>>>>>>>>>>> (k,null) would be emitted out of the join. In this case we
> >> >> don't
> >> >>>>>>>>> need to
> >> >>>>>>>>>>>>> send another null.
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>> Adam
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>> On Wed, Jun 26, 2019 at 11:53 AM John Roesler <
> >> >>>>>>> john@confluent.io>
> >> >>>>>>>>>>>> wrote:
> >> >>>>>>>>>>>>>
> >> >>>>>>>>>>>>>> Hi Adam,
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>>> Thanks for the proposed revision to your KIP
> >> >>>>>>>>>>>>>> (
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>
> >> >>>>>>>>>
> >> >>>>>>>
> >> >>>>
> >> >>
> >> https://cwiki.apache.org/confluence/pages/diffpagesbyversion.action?pageId=74684836&selectedPageVersions=77&selectedPageVersions=74
> >> >>>>>>>>>>>>>> )
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>>> in response to the concern pointed out during code review
> >> >>>>>>>>>>>>>> (
> >> >>>>>>> https://github.com/apache/kafka/pull/5527#issuecomment-505137962
> >> >>>>>>>>> )
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>>> We should have a brief discussion thread (here) in the
> >> mailing
> >> >>>>>>>>> list to
> >> >>>>>>>>>>>>>> make sure everyone who wants to gets a chance to consider
> >> the
> >> >>>>>>>>>>>>>> modification to the design.
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>>> Thanks,
> >> >>>>>>>>>>>>>> -John
> >> >>>>>>>>>>>>>>
> >> >>>>>>>>>>>>
> >> >>>>>>>>>>>
> >> >>>>>>>>>
> >> >>>>>>>
> >> >>>>
> >> >>>>
> >> >>>
> >> >>
> >> >>
> >> >
> >>
> >>


Mime
View raw message