@Lars - Let's say i do a future delete, is there a way to "rollback" this delete without major compaction? @Karthik - memstore-ts will just be the time at which kv arrived into memstore and not the real ordering of the operation. Also perhaps the batch should not rely on KV timestamp for reordering the batch but on total ordering of the batch.
I also think that the behavior should be more deterministic. Mikael.S On Wed, Jan 18, 2012 at 11:51 AM, M. C. Srivas <[email protected]> wrote: > On Tue, Jan 17, 2012 at 8:56 PM, lars hofhansl <[email protected]> > wrote: > > > The memstoreTS is used for visibility during an intra-row transaction. > > Are you proposing to do this only if the deletes/puts did not use the > > current time? > > > > The ability to define timestamps for all operations is crucial to HBase. > > o It ensures that HTable.batch works correctly (which reorders Deletes > > w.r.t. to Puts at the Region Server). > > o It ensures that replication works correctly. > > o many other scenarios > > > > If you do not use application defined timestamp the current time is used > > and everything works as expected. > > If you use application defined timestamps you are asking for a delete to > > be either in the future or the past, and you have to understand what that > > means. > > Maybe we should document the behavior better. > > > > I guess I am saying that I *do* understand the current "delete with TS" > behavior, and I find the current implementation unstable and > non-deterministic. Documenting it more thoroughly does not make it less > quirky or more stable. I propose fixing it along the lines suggested in > option B. Karthik seems to agree. > > > > > > > > -- Lars > > > > > > ----- Original Message ----- > > From: Karthik Ranganathan <[email protected]> > > To: "[email protected]" <[email protected]>; lars hofhansl < > > [email protected]> > > Cc: > > Sent: Tuesday, January 17, 2012 3:27 PM > > Subject: Re: Delete client API. > > > > > > @Srivas - totally agree that B is the correct thing to do. > > > > One way we have talked about implementing this is using the memstore ts. > > Every insert of a KV into the memstore is given a memstore-ts. These are > > persisted only till they are needed (to ensure read atomicity for > > scanners) and then that value is zeroed out on a subsequent compaction > > (saves space). If we retained the memstore-ts even beyond these > > compactions, we could get a deterministic order for the puts and deletes > > (first insert ts < del ts < second insert ts). > > > > Thanks > > Karthik > > > > > > On 1/17/12 2:14 PM, "M. C. Srivas" <[email protected]> wrote: > > > > >On Tue, Jan 17, 2012 at 10:07 AM, lars hofhansl <[email protected]> > > >wrote: > > > > > >> Yeah, it's confusing if one expects it to work like in a relational > > >> database. > > >> You can even do worse. If you by accident place a delete in the future > > >>all > > >> current inserts will be hidden until the next major compaction. :) > > >> I got confused about this myself just recently (see my mail on the > > >> dev-list). > > >> > > >> > > >> In the end this is a pretty powerful feature and core to how HBase > works > > >> (not saying that is not confusing though). > > >> > > >> > > >> If one keeps the following two points in mind it makes more sense: > > >> 1. Delete just sets a tomb stone marker at a specific TS (marking > > >> everything older as deleted). > > >> 2. Everything is versioned, if no version is specified the current > time > > >> (at the regionserver) is used. > > >> > > >> In your example1 below t3 > 6, hence the insert is hidden. > > >> In example2 both delete and insert TS are 6, hence the insert is > hidden. > > >> > > > > > >Lets consider my example2 for a little longer. Sequence of events > > > > > > 1. ins val1 with TS=6 set by client > > > 2. del entire row at TS=6 set by client > > > 3. ins val2 with TS=6 set by client > > > 4. read row > > > > > >The row returns nothing even though the insert at step 3 happened after > > >the > > >delete at step 2. (step 2 masks even future inserts) > > > > > >Now, the same sequence with a compaction thrown in the middle: > > > > > > 1. ins val1 with TS=6 set by client > > > 2. del entire row at TS=6 set by client > > > 3. ---- table is compacted ----- > > > 4. ins val2 with TS=6 set by client > > > 5. read row > > > > > >The row returns val2. (the delete at step2 got lost due to compaction). > > > > > >So we have different results depending upon whether an internal > > >re-organization (like a compaction) happened or not. If we want both > > >sequences to behave exactly the same, then we need to first choose what > is > > >the proper (and deterministic) behavior. > > > > > >A. if we think that the first sequence is the correct one, then the > > >delete > > >at step 2 needs to be preserved forever. > > > > > >or, > > > > > >B. if we think that the second sequence is the correct behavior (ie, a > > >read > > >always produces the same results independent of compaction), then the > > >record needs a second "internal TS" field to allow the RS to distinguish > > >the real sequence of events, and not rely upon the TS field which is > > >settable by the client. > > > > > >My opinion: > > > > > >We should do B. It is normal for someone to write code that says "if > old > > >exists, delete it; add new". A subsequent read should always reliably > > >return "new". > > > > > >The current way of relying on a client-settable TS field to determine > > >causal order results in quirky behavior, and quirky is not good. > > > > > > > > > > > >> Look at these two examples: > > >> > > >> 1. insert Val1 at real time t1 > > >> 2. <del> at real time t2 > t1 > > >> 3. insert Val2 at real time t3 > t2 > > >> > > >> 1. insert Val1 with TS=1 at real time t1 > > >> 2. <del> with TS = 2 at real time t2 > t1 > > >> > > >> 3. insert Val2 with TS = 3 at real time t3 > t2 > > >> > > >> > > >> In both cases Val2 is visible. > > >> > > >> If the your code sets your own timestamps, you better know what you're > > >> doing :) > > >> > > >> Note that my examples below are confusing even if you know how > deletion > > >>in > > >> HBase works. > > >> You have to look at Delete.java to figure out what is happening. > > >> OK, since there were know objections in two days, I will commit my > > >> proposed change in HBASE-5205. > > >> > > >> > > >> -- Lars > > >> > > >> ________________________________ > > >> From: M. C. Srivas <[email protected]> > > >> To: [email protected]; lars hofhansl <[email protected]> > > >> Sent: Tuesday, January 17, 2012 8:13 AM > > >> Subject: Re: Delete client API. > > >> > > >> > > >> Delete seems to be confusing in general. Here are some examples that > > >>make > > >> me scratch my head (key is same in all the examples): > > >> > > >> Example1: > > >> ---------------- > > >> 1. insert Val3 with TS=3 at real time t1 > > >> 2. insert Val5 with TS=5 at real time t2 > t1 > > >> 3. <del> at real time t3 > t2 > > >> 4. insert Val6 with TS=6 at real time t4 > t3 > > >> > > >> What does a read return? (I would expect Val6, since it was done > > >>last). > > >> But depending upon whether compaction happened or not between steps 3 > > >>and > > >> 4, I get either Val6 or nothing. > > >> > > >> Example 2: > > >> ----------------- > > >> 1. insert Val3 with TS=3 at real time t1 > > >> 2. insert Val5 with TS=5 at real time t2 > t1 > > >> 3. <del> TS=6 at real time t3 > t2 > > >> 4. insert Val6 with TS=6 at real time t4 > t3 > > >> > > >> Note the difference in step 3 is this time a TS was specified by the > > >> client. > > >> > > >> What does a read return? Again, I expect Val6 to be returned. But > > >> depending upon what's going on, I seem to get either Val5 or Val6. > > >> > > >> > > >> > > >> > > >> > > >> On Sun, Jan 15, 2012 at 7:21 PM, lars hofhansl <[email protected]> > > >> wrote: > > >> > > >> There are some confusing parts about the Delete client API: > > >> >1. calling deleteFamily removes all prior column or columns markers > > >> without checking the TS. > > >> >2. delete{Column|Columns|Family} do not use the timestamp passed to > > >> Delete at construction time, but instead default to LATEST_TIMESTAMP. > > >> > > > >> > Delete d = new Delete(R,T); > > >> > d.deleteFamily(CF); > > >> > > > >> >Does not do what you expect (won't use T for the family delete, but > > >> rather the current time). > > >> > > > >> >Neither does > > >> > d.deleteColumns(CF, C1, T2); > > >> > d.deleteFamily(CF, T1); // T1 < T2 > > >> > > > >> > > > >> >(the columns marker will be removed) > > >> > > > >> > > > >> >#1 prevents Delete from adding a family marker F for time T1 and a > > >> column/columns marker for columns of F at T2 even if T2 > T1. > > >> >#2 is just unexpected and different from what Put is doing. > > >> > > > >> >In HBASE-5205 I propose a simple patch to fix this. > > >> > > > >> >Since this is a (slight) API change, please provide feed back. > > >> > > > >> >Thanks. > > >> > > > >> >-- Lars > > >> > > > >> > > > >> > > > -- Mikael.S
