couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adam Kocoloski <kocol...@apache.org>
Subject Re: # [DISCUSS] : things we need to solve/decide : storage of edit conflicts
Date Fri, 08 Feb 2019 16:22:39 GMT
Totally unacceptable! ;)  In fact some key bits of that model got dispersed into at least two
separate emails so you’re likely not the only one. I’ll restate here:

The size limits in FoundationDB preclude us from storing the entire key tree as a single value;
in pathological situations the tree could exceed 100KB. Rather, I think it would make sense
to store each edit *branch* as a separate KV. We stem the branch long before it hits the value
size limit, and in the happy case of no edit conflicts this means we store the edit history
metadata in a single KV. It also means that we can apply an interactive edit without retrieving
the entire conflicted revision tree; we need only retrieve and modify the single branch against
which the edit is being applied. The downside is that we may duplicate historical revision
identifiers shared by multiple edit branches, but I think this is a worthwhile tradeoff.

I’d also ensure that a document update only needs to read the edit branch KV against which
the update is being applied, and it can read that branch immediately knowing only the content
of the edit that is being attempted (i.e., it does not need to read the current version of
the document itself).

I think we achieve these goals with a separate subspace (maybe “_meta”?) for the revision
trees, with keys and values that look like

(“_meta”, DocID, NotDeleted, RevPosition, RevHash) = (VersionstampForCurrentRev, [ParentRev,
GrandparentRev, …])

Some notes:

- including IsDeleted ensures that we can efficiently accept the case where we upload a new
document with the same ID where all previous edit branches have been deleted; i.e. we can
construct a key selector which automatically tells us there are no deleted=false edit branches
- access to VersionstampForCurrentRev ensures we can clear the old entry in the by_seq space
during the update
- i need to remind myself how we handle an edit attempt which supplies a _rev representing
a deleted leaf. Do we fail that as a conflict? That would be the natural thing to do here,
otherwise we’re forced to check both deleted=false and deleted=true keys
- the keys can be made to naturally sort so that the winning revision sorts last, but I don’t
believe that’s a requirement here like it is for the actual document data space

Cheers, Adam

> On Feb 8, 2019, at 8:59 AM, Paul Davis <paul.joseph.davis@gmail.com> wrote:
> 
>> I’m relatively happy with the revision history data model at this point.
> 
> I forgot to make a note, but which of the various models are you
> referring to by "revision history data model". There's been so many
> without firm names that my brain is having a hard time parsing that
> one.
> 
> On Thu, Feb 7, 2019 at 9:35 PM Adam Kocoloski <kocolosk@apache.org> wrote:
>> 
>> Bob, Garren, Jan - heard you loud and clear, K.I.S.S. I do think it’s a bit “simplistic"
to exclusively choose simplicity over performance and storage density. We’re (re)building
a database here, one that has some users with pretty demanding performance and scalability
requirements. And yes, we should certainly be testing and measuring. Kyle and team are setting
up infrastructure in IBM land to help with that now, but I also believe we can design the
data model and architecture with a basic performance model of FoundationDB in mind:
>> 
>> - reads cost 1ms
>> - short range reads are the same cost as a single lookup
>> - reads of independent parts of the keyspace can be parallelized for cheap
>> - writes are zero-cost until commit time
>> 
>> We ought to be able to use these assumptions to drive some decisions about data models
ahead of any end-to-end performance test.
>> 
>> If there are specific elements of the edit conflicts management where you think greater
simplicity is warranted, let’s get those called out. Ilya noted (correctly, in my opinion)
that the term sharing stuff is one of those items. It’s relatively complex, potentially
a performance hit, and only saves on storage density in the corner case of lots of edit conflicts.
That’s a good one to drop.
>> 
>> I’m relatively happy with the revision history data model at this point. Hopefully
folks find it easy to grok, and it’s efficient for both reads and writes. It costs some
extra storage for conflict revisions compared to the current tree representation (up to 16K
per edit branch, with default _revs_limit) but knowing what we know about the performance
death spiral for wide revision trees today I’ll happily make a storage vs. performance tradeoff
here :)
>> 
>> Setting the shared term approach aside, I’ve still been mulling over the key structure
for the actual document data:
>> 
>> -  I thought about trying to construct a special _conflicts subspace, but I don’t
like that approach because the choice of a “winning" revision can flip back and forth very
quickly with concurrent writers to different edit branches. I think we really want to have
a way for revisions to naturally sort themselves so the winner is the first or last revision
in a list.
>> 
>> - Assuming we’re using key paths of the form (docid, revision-ish, path, to, field),
the goal here is to find an efficient way to get the last key with prefix “docid” (assuming
winner sorts last), and then all the keys that share the same (docid, revision-ish) prefix
as that one. I see two possible approaches so far, neither perfect:
>> 
>> Option 1: Execute a get_key() operation with a key selector that asks for the last
key less than “docid\xFF” (again assuming winner sorts last), and then do a get_range_startswith()
request setting the streaming mode to “want_all” and the prefix to the docid plus whatever
revision-ish we found from the get_key() request. This is two roundtrips instead of one, but
it always retrieves exactly the right set of keys, and the second step is executed as fast
as possible.
>> 
>> Option 2: Jump straight to get_range_startswith() request using only “docid”
as the prefix, then cancel the iteration once we reach a revision not equal to the first one
we see. We might transfer too much data, or we might end up doing multiple roundtrips if the
default “iterator” streaming mode sends too little data to start (I haven’t checked
what the default iteration block is there), but in the typical case of zero edit conflicts
we have a good chance of retrieving the full document in one roundtrip.
>> 
>> I don’t have a good sense of which option wins out here from a performance perspective,
but they’re both operating on the same data model so easy enough to test the alternatives.
The important bit is getting the revision-ish things to sort correctly. I think we can do
that by generating something like
>> 
>> revision-ish = NotDeleted/1bit : RevPos : RevHash
>> 
>> with some suitable order-preserving encoding on the RevPos integer.
>> 
>> Apologies for the long email. Happy for any comments, either here or over on IRC.
Cheers,
>> 
>> Adam
>> 
>>> On Feb 7, 2019, at 4:52 PM, Robert Newson <rnewson@apache.org> wrote:
>>> 
>>> I think we should choose simple. We can then see if performance is too low or
storage overhead too high and then see what we can do about it.
>>> 
>>> B.
>>> 
>>> --
>>> Robert Samuel Newson
>>> rnewson@apache.org
>>> 
>>> On Thu, 7 Feb 2019, at 20:36, Ilya Khlopotov wrote:
>>>> We cannot do simple thing if we want to support sharing of JSON terms. I
>>>> think if we want the simplest path we should move sharing out of the
>>>> scope. The problem with sharing is we need to know the location of
>>>> shared terms when we do write. This means that we have to read full
>>>> document on every write. There might be tricks to replace full document
>>>> read with some sort of hierarchical signature or sketch of a document.
>>>> However these tricks do not fall into simplest solution category. We
>>>> need to choose the design goals:
>>>> - simple
>>>> - performance
>>>> - reduced storage overhead
>>>> 
>>>> best regards,
>>>> iilyak
>>>> 
>>>> On 2019/02/07 12:45:34, Garren Smith <garren@apache.org> wrote:
>>>>> I’m also in favor of keeping it really simple and then testing and
>>>>> measuring it.
>>>>> 
>>>>> What is the best way to measure that we have something that works? I’m
not
>>>>> sure just relying on our current tests will prove that? Should we define
>>>>> and build some more complex situations e.g docs with lots of conflicts
or
>>>>> docs with wide revisions and make sure we can solve for those?
>>>>> 
>>>>> On Thu, Feb 7, 2019 at 12:33 PM Jan Lehnardt <jan@apache.org> wrote:
>>>>> 
>>>>>> I’m also very much in favour with starting with the simplest thing
that
>>>>>> can possibly work and doesn’t go against the advertised best practices
of
>>>>>> FoundationDB. Let’s get that going and get a feel for how it all
works
>>>>>> together, before trying to optimise things we can’t measure yet.
>>>>>> 
>>>>>> Best
>>>>>> Jan
>>>>>> —
>>>>>> 
>>>>>>> On 6. Feb 2019, at 16:58, Robert Samuel Newson <rnewson@apache.org>
>>>>>> wrote:
>>>>>>> 
>>>>>>> Hi,
>>>>>>> 
>>>>>>> With the Redwood storage engine under development and with prefix
>>>>>> elision part of its design, I don’t think we should get too hung
up on
>>>>>> adding complications and indirections in the key space just yet.
We haven’t
>>>>>> written a line of code or run any tests, this is premature optimisation.
>>>>>>> 
>>>>>>> I’d like to focus on the simplest solution that yields all
required
>>>>>> properties. We can embellish later (if warranted).
>>>>>>> 
>>>>>>> I am intrigued by all the ideas that might allow us cheaper inserts
and
>>>>>> updates than the current code where there are multiple edit branches
in the
>>>>>> stored document.
>>>>>>> 
>>>>>>> B.
>>>>>>> 
>>>>>>>> On 6 Feb 2019, at 02:18, Ilya Khlopotov <iilyak@apache.org>
wrote:
>>>>>>>> 
>>>>>>>> While reading Adam's proposal I came to realize that: we
don't have to
>>>>>> calculate winning revision at read time.
>>>>>>>> Since FDB's transactions are atomic we can calculate it when
we write.
>>>>>> This means we can just write latest values into separate range. This
makes
>>>>>> lookup of latest version fast.
>>>>>>>> Another realization is if we want to share values for some
json paths
>>>>>> we would have to introduce a level of indirection.
>>>>>>>> Bellow is the data model inspired by Adam's idea to share
json_paths.
>>>>>> In this model the json_path is stored in the revision where it was
first
>>>>>> added (we call that revision an owner of a json_path). The values
for
>>>>>> json_path key can be scalar values, parts of scalar values or pointers
to
>>>>>> owner location.
>>>>>>>> The below snippets are sketches of transactions.
>>>>>>>> The transactions will include updates to other keys as needed
>>>>>> (`external_size`, `by_seq` and so on).  The revision tree management
is not
>>>>>> covered yet.
>>>>>>>> The `rev -> vsn` indirection is not strictly required.
It is added
>>>>>> because it saves some space since `rev` is a long string and `vsn`
is FDB
>>>>>> versionstamp of fixed size.
>>>>>>>> 
>>>>>>>> - `{NS} / {docid} / _by_rev / {rev} = vsn`
>>>>>>>> - `{NS} / {docid} / _used_by / {json_path} / {another_vsn}
= NIL`
>>>>>>>> - `{NS} / {docid} / _data / {json_path} = latest_value |
part`
>>>>>>>> - `{NS} / {docid} / {vsn} / _data / {json_path} = value |
part |
>>>>>> {another_vsn}`
>>>>>>>> 
>>>>>>>> ```
>>>>>>>> write(txn, doc_id, prev_rev, json):
>>>>>>>> txn.add_write_conflict_key("{NS} / {doc_id} / _rev")
>>>>>>>> rev = generate_new_rev()
>>>>>>>> txn["{NS} / {docid} / _by_rev / {rev}"] = vsn
>>>>>>>> for every json_path in flattened json
>>>>>>>>  - {NS} / {docid} / _used_by / {json_path} / {another_vsn}
= NIL
>>>>>>>>  if rev is HEAD:
>>>>>>>>    # this range contains values for all json paths for the
latest
>>>>>> revision (read optimization)
>>>>>>>>    - {NS} / {docid} / _data / {json_path} = latest_value
| part
>>>>>>>>  - {NS} / {docid} / {vsn} / _data / {json_path} = value |
part |
>>>>>> {another_vsn}
>>>>>>>> txn["{NS} / {doc_id} / _rev"] = rev
>>>>>>>> 
>>>>>>>> get_current(txn, doc_id):
>>>>>>>> # there is no sharing of json_paths in this range (read optimization)
>>>>>>>> txn.get_range("{NS} / {docid} / _data / 0x00", "{NS} / {docid}
/ _data
>>>>>> / 0xFF" )
>>>>>>>> 
>>>>>>>> get_revision(txn, doc_id, rev):
>>>>>>>> vsn = txn["{NS} / {docid} / _by_rev / {rev}"]
>>>>>>>> json_paths = txn.get_range("{NS} / {vsn} / {docid} / _data
/ 0x00",
>>>>>> "{NS} / {vsn} / {docid} / _data / 0xFF" )
>>>>>>>> for every json_path in json_paths:
>>>>>>>> if value has type vsn:
>>>>>>>>    another_vsn = value
>>>>>>>>       value = txn["{NS} / {docid} / {another_vsn} / _data
/
>>>>>> {json_path}"]
>>>>>>>> result[json_path] = value
>>>>>>>> 
>>>>>>>> delete_revision(txn, doc_id, rev):
>>>>>>>> vsn = txn["{NS} / {docid} / _by_rev / {rev}"]
>>>>>>>> json_paths = txn.get_range("{NS} / {vsn} / {docid} / _data
/ 0x00",
>>>>>> "{NS} / {vsn} / {docid} / _data / 0xFF" )
>>>>>>>> for every json_path in json_paths:
>>>>>>>> if value has type vsn:
>>>>>>>>   # remove reference to deleted revision from the owner
>>>>>>>>    del txn[{NS} / {docid} / _used_by / {json_path} / {vsn}]
>>>>>>>> # check if deleted revision of json_path is not used by anything
else
>>>>>>>> if txn.get_range("{NS} / {docid} / _used_by / {json_path}
/ {vsn}",
>>>>>> limit=1) == []:
>>>>>>>>    del txn["{NS} / {docid} / {vsn} / _data / {json_path}"]
>>>>>>>> if vsn is HEAD:
>>>>>>>>    copy range for winning revision into "{NS} / {docid} /
_data /
>>>>>> {json_path}"
>>>>>>>> ```
>>>>>>>> 
>>>>>>>> best regards,
>>>>>>>> iilyak
>>>>>>>> 
>>>>>>>> On 2019/02/04 23:22:09, Adam Kocoloski <kocolosk@apache.org>
wrote:
>>>>>>>>> I think it’s fine to start a focused discussion here
as it might help
>>>>>> inform some of the broader debate over in that thread.
>>>>>>>>> 
>>>>>>>>> As a reminder, today CouchDB writes the entire body of
each document
>>>>>> revision on disk as a separate blob. Edit conflicts that have common
fields
>>>>>> between them do not share any storage on disk. The revision tree
is encoded
>>>>>> into a compact format and a copy of it is stored directly in both
the by_id
>>>>>> tree and the by_seq tree. Each leaf entry in the revision tree contain
a
>>>>>> pointer to the position of the associated doc revision on disk.
>>>>>>>>> 
>>>>>>>>> As a further reminder, CouchDB 2.x clusters can generate
edit conflict
>>>>>> revisions just from multiple clients concurrently updating the same
>>>>>> document in a single cluster. This won’t happen when FoundationDB
is
>>>>>> running under the hood, but users who deploy multiple CouchDB or
PouchDB
>>>>>> servers and replicate between them can of course still produce conflicts
>>>>>> just like they could in CouchDB 1.x, so we need a solution.
>>>>>>>>> 
>>>>>>>>> Let’s consider the two sub-topics separately: 1) storage
of edit
>>>>>> conflict bodies and 2) revision trees
>>>>>>>>> 
>>>>>>>>> ## Edit Conflict Storage
>>>>>>>>> 
>>>>>>>>> The simplest possible solution would be to store each
document
>>>>>> revision separately, like we do today. We could store document bodies
with
>>>>>> (“docid”, “revid”) as the key prefix, and each transaction
could clear the
>>>>>> key range associated with the base revision against which the edit
is being
>>>>>> attempted. This would work, but I think we can try to be a bit more
clever
>>>>>> and save on storage space given that we’re splitting JSON documents
into
>>>>>> multiple KV pairs.
>>>>>>>>> 
>>>>>>>>> One thought I’d had is to introduce a special enum
Value which
>>>>>> indicates that the subtree “beneath” the given Key is in conflict.
For
>>>>>> example, consider the documents
>>>>>>>>> 
>>>>>>>>> {
>>>>>>>>> “_id”: “foo”,
>>>>>>>>> “_rev”: “1-abc”,
>>>>>>>>> “owner”: “alice”,
>>>>>>>>> “active”: true
>>>>>>>>> }
>>>>>>>>> 
>>>>>>>>> and
>>>>>>>>> 
>>>>>>>>> {
>>>>>>>>> “_id”: “foo”,
>>>>>>>>> “_rev”: “1-def”,
>>>>>>>>> “owner”: “bob”,
>>>>>>>>> “active”: true
>>>>>>>>> }
>>>>>>>>> 
>>>>>>>>> We could represent these using the following set of KVs:
>>>>>>>>> 
>>>>>>>>> (“foo”, “active”) = true
>>>>>>>>> (“foo”, “owner”) = kCONFLICT
>>>>>>>>> (“foo”, “owner”, “1-abc”) = “alice”
>>>>>>>>> (“foo”, “owner”, “1-def”) = “bob”
>>>>>>>>> 
>>>>>>>>> This approach also extends to conflicts where the two
versions have
>>>>>> different data types. Consider a more complicated example where bob
dropped
>>>>>> the “active” field and changed the “owner” field to an object:
>>>>>>>>> 
>>>>>>>>> {
>>>>>>>>> “_id”: “foo”,
>>>>>>>>> “_rev”: “1-def”,
>>>>>>>>> “owner”: {
>>>>>>>>> “name”: “bob”,
>>>>>>>>> “email”: “bob@example.com"
>>>>>>>>> }
>>>>>>>>> }
>>>>>>>>> 
>>>>>>>>> Now the set of KVs for “foo” looks like this (note
that a missing
>>>>>> field needs to be handled explicitly):
>>>>>>>>> 
>>>>>>>>> (“foo”, “active”) = kCONFLICT
>>>>>>>>> (“foo”, “active”, “1-abc”) = true
>>>>>>>>> (“foo”, “active”, “1-def”) = kMISSING
>>>>>>>>> (“foo”, “owner”) = kCONFLICT
>>>>>>>>> (“foo”, “owner”, “1-abc”) = “alice”
>>>>>>>>> (“foo”, “owner”, “1-def”, “name”) = “bob”
>>>>>>>>> (“foo”, “owner”, “1-def”, “email”) =
“bob@example.com”
>>>>>>>>> 
>>>>>>>>> I like this approach for the common case where documents
share most of
>>>>>> their data in common but have a conflict in a very specific field
or set of
>>>>>> fields.
>>>>>>>>> 
>>>>>>>>> I’ve encountered one important downside, though: an
edit that
>>>>>> replicates in and conflicts with the entire document can cause a
bit of a
>>>>>> data explosion. Consider a case where I have 10 conflicting versions
of a
>>>>>> 100KB document, but the conflicts are all related to a single scalar
value.
>>>>>> Now I replicate in an empty document, and suddenly I have a kCONFLICT
at
>>>>>> the root. In this model I now need to list out every path of every
one of
>>>>>> the 10 existing revisions and I end up with a 1MB update. Yuck. That’s
>>>>>> technically no worse in the end state than the “zero sharing”
case above,
>>>>>> but one could easily imagine overrunning the transaction size limit
this
>>>>>> way.
>>>>>>>>> 
>>>>>>>>> I suspect there’s a smart path out of this. Maybe the
system detects a
>>>>>> “default” value for each field and uses that instead of writing
out the
>>>>>> value for every revision in a conflicted subtree. Worth some discussion.
>>>>>>>>> 
>>>>>>>>> ## Revision Trees
>>>>>>>>> 
>>>>>>>>> In CouchDB we currently represent revisions as a hash
history tree;
>>>>>> each revision identifier is derived from the content of the revision
>>>>>> including the revision identifier of its parent. Individual edit
branches
>>>>>> are bounded in *length* (I believe the default is 1000 entries),
but the
>>>>>> number of edit branches is technically unbounded.
>>>>>>>>> 
>>>>>>>>> The size limits in FoundationDB preclude us from storing
the entire
>>>>>> key tree as a single value; in pathological situations the tree could
>>>>>> exceed 100KB. Rather, I think it would make sense to store each edit
>>>>>> *branch* as a separate KV. We stem the branch long before it hits
the value
>>>>>> size limit, and in the happy case of no edit conflicts this means
we store
>>>>>> the edit history metadata in a single KV. It also means that we can
apply
>>>>>> an interactive edit without retrieving the entire conflicted revision
tree;
>>>>>> we need only retrieve and modify the single branch against which
the edit
>>>>>> is being applied. The downside is that we duplicate historical revision
>>>>>> identifiers shared by multiple edit branches, but I think this is
a
>>>>>> worthwhile tradeoff.
>>>>>>>>> 
>>>>>>>>> I would furthermore try to structure the keys so that
it is possible
>>>>>> to retrieve the “winning” revision in a single limit=1 range
query. Ideally
>>>>>> I’d like to proide the following properties:
>>>>>>>>> 
>>>>>>>>> 1) a document read does not need to retrieve the revision
tree at all,
>>>>>> just the winning revision identifier (which would be stored with
the rest
>>>>>> of the doc)
>>>>>>>>> 2) a document update only needs to read the edit branch
of the
>>>>>> revision tree against which the update is being applied, and it can
read
>>>>>> that branch immediately knowing only the content of the edit that
is being
>>>>>> attempted (i.e., it does not need to read the current version of
the
>>>>>> document itself).
>>>>>>>>> 
>>>>>>>>> So, I’d propose a separate subspace (maybe “_meta”?)
for the revision
>>>>>> trees, with keys and values that look like
>>>>>>>>> 
>>>>>>>>> (“_meta”, DocID, IsDeleted, RevPosition, RevHash)
= [ParentRev,
>>>>>> GrandparentRev, …]
>>>>>>>>> 
>>>>>>>>> The inclusion of IsDeleted, RevPosition and RevHash in
the key should
>>>>>> be sufficient (with the right encoding) to create a range query that
>>>>>> automatically selects the “winner” according to CouchDB’s arcane
rules,
>>>>>> which are something like
>>>>>>>>> 
>>>>>>>>> 1) deleted=false beats deleted=true
>>>>>>>>> 2) longer paths (i.e. higher RevPosition) beat shorter
ones
>>>>>>>>> 3) RevHashes with larger binary values beat ones with
smaller values
>>>>>>>>> 
>>>>>>>>> ===========
>>>>>>>>> 
>>>>>>>>> OK, that’s all on this topic from me for now. I think
this is a
>>>>>> particularly exciting area where we start to see the dividends of
splitting
>>>>>> up data into multiple KV pairs in FoundationDB :) Cheers,
>>>>>>>>> 
>>>>>>>>> Adam
>>>>>>>>> 
>>>>>>>>> 
>>>>>>>>>> On Feb 4, 2019, at 2:41 PM, Robert Newson <rnewson@apache.org>
wrote:
>>>>>>>>>> 
>>>>>>>>>> This one is quite tightly coupled to the other thread
on data model,
>>>>>> should we start much conversation here before that one gets closer
to a
>>>>>> solution?
>>>>>>>>>> 
>>>>>>>>>> --
>>>>>>>>>> Robert Samuel Newson
>>>>>>>>>> rnewson@apache.org
>>>>>>>>>> 
>>>>>>>>>> On Mon, 4 Feb 2019, at 19:25, Ilya Khlopotov wrote:
>>>>>>>>>>> This is a beginning of a discussion thread about
storage of edit
>>>>>>>>>>> conflicts and everything which relates to revisions.
>>>>>>>>>>> 
>>>>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> 
>>>>>>> 
>>>>>> 
>>>>>> --
>>>>>> Professional Support for Apache CouchDB:
>>>>>> https://neighbourhood.ie/couchdb-support/
>>>>>> 
>>>>>> 
>>>>> 
>> 


Mime
View raw message