This patch adds several more tests for Alternator's UpdateItem operation.
These tests verify a few simple cases that, surprisingly, never had test
coverage. The new tests pass (on both DynamoDB and Alternator) so did not
expose any bug.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#11025
If the compaction_descriptor returned by `time_window_compaction_strategy::get_sstables_for_compaction`
is marked with `has_only_fully_expired::yes` it should always be compacted
since `time_window_compaction_strategy::get_sstables_for_compaction` is not idempotent.
It sets `_last_expired_check` and if compaction is postponed and retried before
`expired_sstable_check_frequency` has passed, it will not look for those fully-expired
sstables again. Plus, compacting them is the cheapest possible as it does not require
reading anything, just deleting the input sstables, so there's no reason not postpone it.
Also, extend `max_ongoing_compaction_test` to test serialization of compaction jobs with the same weight.
Fixes#10989Closes#10990
* github.com:scylladb/scylla:
compaction_manager: always register descriptor with fully expired sstables for compaction
test: max_ongoing_compaction_test: test serialization of regular compaction with same weight
test: max_ongoing_compaction_test: reindent refactored code
test: max_ongoing_compaction_test: define compact_all_tables lambda
test: max_ongoing_compaction_test: refactor make_table_with_single_fully_expired_sstable
test: max_ongoing_compaction_test: reduce number of tables
This series adds the infrastructure needed for testing user permissions, like the ability to create temporary roles and CQL sessions which log in as different users, and a few initial test cases for granting and revoking permissions.
Closes#10998
* github.com:scylladb/scylla:
cql-pytest: add a case for granting/revoking data permissions
cql-pytest: add new_user and new_session utils
cql-pytest: speed up permissions refresh period for tests
Said parameter is a convenience so downstream consumers of the
mutation compactors don't have to check the `bool is_live` already
passed to them. This convenience however causes a template parameter and
additional logic for the compactor. As the most prominent of these
consumers (the query result builder) will soon have to switch to
`emit_only_live_rows::no` for other reasons anyway (it will want to count
tombstones), we take the opportunity to switch everybody to ::no. This
can be done with very little additional complexity to these consumers --
basically an additional if or two. With everybody using the `::no` variant
of the compactor, we can remove this template parameter and the logic
associated with it altogether.
Closes#10931
* github.com:scylladb/scylla:
multishard_mutation_query: remove now pointless compact_for_result_state typedef
mutation_compactor: remove only-live related logic
mutation_compactor: remove emit_only_live_rows template parameter
mutation_compactor: remove unused compact_mutation_state::parameters
querier: remove {data,mutation}_querier aliases
querier: remove now pointless emit_only_live_rows template parameter
tree: use emit_only_live_rows::no
querier: querier_cache: de-override insert() methods
If the compaction_descriptor returned by time_window_compaction_strategy::get_sstables_for_compaction
is marked with has_only_fully_expired::yes it should always be compacted
since time_window_compaction_strategy::get_sstables_for_compaction is not idempotent.
It sets _last_expired_check and if compaction is postponed and retried before
expired_sstable_check_frequency has passed, it will not look for those fully-expired
sstables again. Plus, compacting them is the cheapest possible as it does not require
reading anything, just deleting the input sstables, so there's no reason not postpone it.
Fixes#10989
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
To test both expired and non-expired sstables scenarios
we need to pass this helper function the expected number
of sstables before compaction and after compaction.
When compaction a set of fully-expired sstables,
we expect none to remain, while when the set of sstables
is not fully expired, we'll expect 1 output sstable
after compaction.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
So we can use the lower-level build blocks to
test compaction serialization of both fully-expired
and non-fully-expired sstables scenarios in the following patches.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Now that we use emit_only_live_rows::no everywhere we can remove this
template parameters. Only the template parameter is removed, the
internal logic around it is left in place (will be removed in a next
patch), by hard-wiring `only_live()`.
emit_only_live_rows is a convenience so downstream consumers of the
mutation compactors don't have to check the `bool is_live` already
passed to them. This convenience however causes a template parameter and
additional logic for the compactor. As the most prominent of these
consumers (the query result builder) will soon have to switch to
emit_only_live_rows::no for other reasons anyway (it will want to count
tombstones), we take the opportunity to switch everybody to ::no. This
can be done with very little additional complexity to these consumer --
basically an additional if or two.
This prepares the ground for removing this template parameter and the
associate logic from the compactor.
Soon, the currently two distinct types of queriers will be merged, as
the template parameter differentiating them will be gone. This will make
using type based overload for insert() impossible, as 2 out of the 3
types will be the same. Use different names instead.
This PR introduces improvements to `expr::to_restriction` and prepares the validation part for restriction classes removal.
`expr::to_restriction` is currently used to take a restriction from the WHERE clause, prepare it, perform some validation checks and finally convert it to an instance of the restriction class.
Soon we will get rid of the restriction class.
In preparation for that `expr::to_restriction` is split into two independent parts:
* The part that prepares and validates a binary_operator
* The part that converts a binary_operator to restriction
Thanks to this split getting rid of restriction class will be painless, we will just stop using the second part.
`to_restriction.cc` is replaced by `restrictions.hh/cc`. In the future we can put all the restriction expressions code there to avoid clutter in `expression.hh/cc`.
This change made it much easier to fix#10631, so I did that as well.
Fixes: #10631Closes#10979
* github.com:scylladb/scylla:
cql-pytest: Test that IS NOT only accepts NULL
cql-pytest: Enable testInvalidCollectionNonEQRelation
cql3: Move single element IN restrictions handling
cql3: Check for disallowed operators early
cql3: Simplify adding restrictions
cql3: Reorganize to_restriction code
cql3: Fix IS NOT NULL check in to_restriction
cql3: Swap order of arguments in error message
bytes_ostream is an incremental builder for a discontiguous byte container.
managed_bytes is a non-incremental (size must be known up front) byte
container, that is also compatible with LSA. So far, conversion between
them involves copying. This is unfortunate, since query_result is generated
as a bytes_ostream, but is later converted to managed_bytes (today, this
is done in cql3::expr::get_non_pk_values() and
compound_view_wrapper::explode(). If the two types could be made compatible,
we could use managed_bytes_view instead of creating new objects and avoid
a copy. It's also nicer to have one less vocabulary type.
This patch makes bytes_ostream use managed_bytes' internal representation
(blob_storage instead of bytes_ostream::chunk) and provides a conversion
to managed_bytes. All bytes_ostream users are left in place, but the goal
is to make bytes_ostream a write-only type with the only observer a conversion
to managed_bytes.
It turns out to be relatively simple. The internal representations were
already similar. I made blob_storage::ref_type self-initializing to
reduce churn (good practice anyway) and added a private constructor
to managed_bytes for the conversion.
Note that bytes_ostream can only be used to construct a non-LSA managed_bytes,
but LSA uses of managed_bytes are very strictly controlled (the entry
points to memtable and cache) so that's not a problem.
A unit test is added.
Closes#10986
After acquiring the _compaction_state write lock,
select all sstables using get_candidates and register them
as compacting, then unlock the _compaction_state lock
to let regular compaction run in parallel.
Also, run major compaction in maintenance scheduling group.
We should separate the scheduling groups used for major compaction
from the the regular compaction scheduling group so that
the latter can be affected by the backlog tracker in case
backlog accumulates during a long running major compaction.
Fixes#10961Closes#10984
* github.com:scylladb/scylla:
compaction_manager: major_compaction_task: run in maintenance scheduling groupt
compaction_manager: allow regular compaction to run in parallel to major
The IS_NOT operator can only be used during materialized view creation
and it can only be used to express IS NOT NULL.
Trying to write something like IS NOT 42 should cause an error.
Signed-off-by: Jan Ciolek <jan.ciolek@scylladb.com>
This PR gets rid of exception throws/rethrows on the replica side for writes and single-partition reads. This goal is achieved without using `boost::outcome` but rather by replacing the parts of the code which throw with appropriate seastar idioms and by introducing two helper functions:
1.`try_catch` allows to inspect the type and value behind an `std::exception_ptr`. When libstdc++ is used, this function does not need to throw the exception and avoids the very costly unwind process. This based on the "How to catch an exception_ptr without even try-ing" proposal mentioned in https://github.com/scylladb/scylla/issues/10260.
This function allows to replace the current `try..catch` chains which inspect the exception type and account it in the metrics.
Example:
```c++
// Before
try {
std::rethrow_exception(eptr);
} catch (std::runtime_exception& ex) {
// 1
} catch (...) {
// 2
}
// After
if (auto* ex = try_catch<std::runtime_exception>(eptr)) {
// 1
} else {
// 2
}
```
2. `make_nested_exception_ptr` which is meant to be a replacement for `std::throw_with_nested`. Unlike the original function, it does not require an exception being currently thrown and does not throw itself - instead, it takes the nested exception as an `std::exception_ptr` and produces another `std::exception_ptr` itself.
Apart from the above, seastar idioms such as `make_exception_future`, `co_await as_future`, `co_return coroutine::exception()` are used to propagate exceptions without throwing. This brings the number of exception throws to zero for single partition reads and writes (tested with scylla-bench, --mode=read and --mode=write).
Results from `perf_simple_query`:
```
Before (719724e4df):
Writes:
Normal:
127841.40 tps ( 56.2 allocs/op, 13.2 tasks/op, 50042 insns/op, 0 errors)
Timeouts:
94770.81 tps ( 53.1 allocs/op, 5.1 tasks/op, 78678 insns/op, 1000000 errors)
Reads:
Normal:
138902.31 tps ( 65.1 allocs/op, 12.1 tasks/op, 43106 insns/op, 0 errors)
Timeouts:
62447.01 tps ( 49.7 allocs/op, 12.1 tasks/op, 135984 insns/op, 936846 errors)
After (d8ac4c02bfb7786dc9ed30d2db3b99df09bf448f):
Writes:
Normal:
127359.12 tps ( 56.2 allocs/op, 13.2 tasks/op, 49782 insns/op, 0 errors)
Timeouts:
163068.38 tps ( 52.1 allocs/op, 5.1 tasks/op, 40615 insns/op, 1000000 errors)
Reads:
Normal:
151221.15 tps ( 65.1 allocs/op, 12.1 tasks/op, 43028 insns/op, 0 errors)
Timeouts:
192094.11 tps ( 41.2 allocs/op, 12.1 tasks/op, 33403 insns/op, 960604 errors)
```
Closes#10368
* github.com:scylladb/scylla:
database: avoid rethrows when handling exceptions from commitlog
database: convert throw_commitlog_add_error to use make_nested_exception_ptr
utils: add make_nested_exception_ptr
storage_proxy: don't rethrow when inspecting replica exceptions on write path
database: don't rethrow rate_limit_exception
storage_proxy: don't rethrow the exception in abstract_read_resolver::error
utils/exceptions.cc: don't rethrow in is_timeout_exception
utils/exceptions: add try_catch
utils: add abi/eh_ia64.hh
storage_proxy: don't rethrow exceptions from replicas when accounting read stats
message: get rid of throws in send_message{,_timeout,_abortable}
database/{query,query_mutations}: don't rethrow read semaphore exceptions
The default refresh period for permissions in both Scylla and Cassandra
is 2 seconds, which is usually perfectly fine for production
environments, but it introduces a significant delay in automatic
test cases. The refresh period is hereby set to 100ms, which allows
test_permissions.py cases to run in around 1s for Scylla instead of
tens of seconds.
Recently we noticed a regression where with certain versions of the fmt
library,
SELECT value FROM system.config WHERE name = 'experimental_features'
returns string numbers, like "5", instead of feature names like "raft".
It turns out that the fmt library keep changing their overload resolution
order when there are several ways to print something. For enum_option<T> we
happen to have to conflicting ways to print it:
1. We have an explicit operator<<.
2. We have an *implicit* convertor to the type held by T.
We were hoping that the operator<< always wins. But in fmt 8.1, there is
special logic that if the type is convertable to an int, this is used
before operator<<()! For experimental_features_t, the type held in it was
an old-style enum, so it is indeed convertible to int.
The solution I used in this patch is to replace the old-style enum
in experimental_features_t by the newer and more recommended "enum class",
which does not have an implicit conversion to int.
I could have fixed it in other ways, but it wouldn't have been much
prettier. For example, dropping the implicit convertor would require
us to change a bunch of switch() statements over enum_option (and
not just experimental_features_t, but other types of enum_option).
Going forward, all uses of enum_option should use "enum class", not
"enum". tri_mode_restriction_t was already using an enum class, and
now so does experimental_features_t. I changed the examples in the
comments to also use "enum class" instead of enum.
This patch also adds to the existing experimental_features test a
check that the feature names are words that are not numbers.
Fixes#11003.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#11004
Fix two cql-pytest that have been "XPASS"ing (unexpectedly passing)
by removing the "xfail" (expecting failure) mark from them:
One test was for an issue that has already been fixed (refs #10081).
The second test was a translated Cassandra test that should never
have failed because it doesn't trigger the issue that supposedly failed
it (that test sets a large value for a non-indexed column, so doesn't
trigger the problem we have with large values in an indexed column).
Closes#11006
When running test/cql-pytest, pytest prints one warning at the end:
/home/nyh/scylla/test/cql-pytest/test_secondary_index.py:82: DeprecationWarning: ResultSet indexing support will be removed in 4.0.
Consider using ResultSet.one() to get a single row.
assert any([index_name in event.description for event in cql.execute(query, trace=True).get_query_trace().events])
So in this patch I do exactly what the warning recommends - use one().
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#11002
Python has deprecated the distutils package. In several places in the
Alternator and Redis test suites, we used distutils.version to check if
the library is new enough for running the test (and skip the test if
it's too old). On new versions of Python, we started getting deprecation
warnings such as:
DeprecationWarning: The distutils package is deprecated and slated for
removal in Python 3.12. Use setuptools or check PEP 632 for potential
alternatives
PEP 632 recommends using package.version instead of distutils.version,
and indeed it works well. After applying this patch, Alternator and
Redis test runs no long end in silly deprecation warnings.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#11007
This new test suite is expected to gather all kinds of permissions
tests - granting, revoking, authorizing, and so on.
Right now it contains a single minimal test which ensures that
the default superuser can be granted applicable permissions,
which they already have anyway.
The test suite added in this pull request will also be useful
when developing #10633 - permissions for UDF/UDA infrastructure.
Closes#10991
* github.com:scylladb/scylla:
cql-pytest: add initial permissions test suite
cql-pytest: enable CassandraAuthorizer for Scylla and Cassandra
This new test suite is expected to gather all kinds of permissions
tests - granting, revoking, authorizing, and so on.
Right now it contains a single minimal test which ensures that
the default superuser can be granted applicable permissions,
which they already have anyway.
In order to be able to test permissions, an authorizer different
than AllowAllAuthorizer (default) must be set.
CassandraAuthorizer is thus enabled - it works on default user/password
pair, so it doesn't introduce any regressions to the test suite.
We should separate the scheduling groups used for major compaction
from the the regular compaction scheduling group so that
the latter can be affected by the backlog tracker in case
backlog accumulates during a long running major compaction.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Leader which ceases to be a leader as a result of a
execute_modify_config cannot wait for a dummy record to be
committed because io_fiber aborts current waiters as soon as it
detects a lost of leadership.
This commit excludes dummy entries from the configuration change
procedure. A special promise is set on io_fiber when it gets a
non-joint configuration, and set_configuration just waits for
the corresponding future instead of a dummy record.
Fixes: #10010Closes#10905
This reverts commit aa8f135f64, reversing
changes made to 9a88bc260c. The patch
causes hangs during flush.
Also reverts parts of 411231da75 that impacted the unit test.
Fixes#10897.
Native types were parsed directly to data_type, where varchar and text were
parsed to utf8_type. To get the name of the type there was a call to
the data_type method thus getting the name of the varchar type returns "text".
To fix this, added new nonterminal type_unreserved_keyword, which parse native
types to their names. It replaced native_or_internal_type in unreserved_function_keyword.
unreserved_function_keyword is also used to parse usernames, keyspace names, index names,
column identifieres, service levels and role names, so this bug was repaired also in them.
Fixes: #10642Closes#10960
The utils::make_nested_exception_ptr function works similar to
std::throw_with_nested, but instead of storing the currently thrown
exception as the nested exception and then immediately throwing the new
exception, it receives the nested exception as an std::exception_ptr and
also returns an std::exception_ptr.
If the standard library supports it, the function does not perform any
throws. Otherwise the fallback logic performs two throws.
Introduces a utility function which allows obtaining a pointer to the
exception data held behind an std::exception_ptr if the data matches the
requested type. It can be used to implement manual but concise
try..catch chains.
The `try_catch` has the best performance when used with libstdc++ as it
uses the stdlib specific functions for simulating a try..catch without
having to actually throw. For other stdlibs, the implementation falls
back to a throw surrounded by an actual try..catch.
Recently a change to Scylla's expression implementation changed the standard
error message copied from Cassandra:
Cannot execute this query as it might involve data filtering and thus
may have unpredictable performance. If you want to execute this query
despite the performance unpredictability, use ALLOW FILTERING
In the special case where the filter is on the partition key, we changed
the message to:
Only EQ and IN relation are supported on the partition key (unless you
use the token() function or allow filtering)
We had a cql-pytest test translated from Cassandra's unit test that checked
the old message, and started to fail. Unfortunately nobody noticed because
a bug in test.py caused it to stop running these translated unit tests.
So in this patch, we trivially fix the test to pass again. Instead of
insisting on the old message, we check jsut for the string "allow
filtering", in lowercase or uppercase. After this patch, the tests
passes as expected on both Scylla and Cassandra.
Refs #10918 (this test failing is one of the failures reported there)
Refs #10962 (test.py stopped running this test)
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#10964
with several Java versions
The test/cql-pytest/run-cassandra script runs our cql-pytest tests against
Cassandra. Today, Cassandra can only run correctly on Java 8 or 11
(see https://issues.apache.org/jira/browse/CASSANDRA-16895) but recent
Linux distributions have switched to newer versions of Java - e.g., on
my Fedora 36 installation, the default "java" is Java 17. Which can't
run Cassandra.
So what I do in this patch is to check if "java" has the right version,
and if it doesn't, it looks at several additional locations if it can
find a Java of the right version. By the way, we are sure that Java 8
must be installed because our install-dependencies.sh installs it.
After this patch, test/cql-pytest/run-cassandra resumes working on
Fedora 36.
Fixes#10946
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#10947
Method reponsible for creating a token of given values is not meant to be
used with empty optionals. Thus, having requested a token of the columns
containing null values resulted with an exception being thrown. This kind
of behaviour was not compatible with the one applied in cassandra.
To fix this, before the computation of a token, it is checked whether
no null value is contained. If any value in the processed vector is null,
null value is returned.
Fixes: #10594Closes#10942
In order to allow our Scylla OSS customers the ability to select a version for their documentation, we are migrating the Scylla docs content to the Scylla OSS repository. This PR covers the following points of the [Migration Plan](https://docs.google.com/document/d/15yBf39j15hgUVvjeuGR4MCbYeArqZrO1ir-z_1Urc6A/edit#):
1. Creates a subdirectory for dev docs: /docs/dev
2. Moves the existing dev doc content in the scylla repo to /docs/dev, but keep Alternator docs in /docs.
3. Flattens the structure in /docs/dev (remove the subfolders).
4. Adds redirects from `scylla.docs.scylladb.com/<version>/<document>` to `https://github.com/scylladb/scylla/blob/master/docs/dev/<document>.md`
5. Excludes publishing docs for /docs/devs.
1. Enter the docs folder with `cd docs`.
2. Run `make redirects`.
3. Enter the docs folder and run `make preview`. The docs should build without warnings.
4. Open http://127.0.0.1:5500 in your browser. You shoul donly see the alternator docs.
5. Open http://127.0.0.1:5500/stable/design-notes/IDL.html in your browser. It should redirect you to https://github.com/scylladb/scylla/blob/master/docs/dev/IDL.md and raise a 404 error since this PR is not merged yet.
6. Surf the `docs/dev` folder. It should have all the scylla project internal docs without subdirectories.
Closes#10873
* github.com:scylladb/scylla:
Update docs/conf.py
Update docs/dev/protocols.md
Update docs/dev/README.md
Update docs/dev/README.md
Update docs/conf.py
Fix broken links
Remove source folder
Add redirections
Move dev docs to docs/dev
After compiling to WASM, UDFs become much larger than the
source code. When they're included in test_wasm.py, it
becomes difficult to navigate in the file. Moving them
to another place does not make understanding the test
scripts harder, because the source code is still included.
This problem will become even more severe when testing
UDFs using WASI.
Signed-off-by: Wojciech Mitros <wojciech.mitros@scylladb.com>
Closes#10934
To call a UDF that is using WASI, we need to properly
configure the wasmtime instance that it will be called
on. The configuration was missing from udf_cache::load(),
so we add it here.
The free function does not return any value, so we should use
a calling method that does not expect any returns.
This patch adds such a method and uses it.
A test that did not pass without this fix and does pass after
is added.
Signed-off-by: Wojciech Mitros <wojciech.mitros@scylladb.com>
Closes#10935
test/scylla-gdb tests Scylla's gdb debugging tools, and cannot work if
Scylla was compiled without debug information (i.e, the "dev" build mode).
In the past, test/scylla-gdb/run detected this case and printed a clear error:
Scylla executable was compiled without debugging information (-g)
so cannot be used to test gdb. Please set SCYLLA environment variable.
Unfortunately, since recently this detection fails, because even when
Scylla is compiled without debug information we link into it a library
(libwasmtime.a) which has *some* debug information. As a result, instead
of one clear error message, we get all scylla-gdb tests running -
and each of them failing separately. This is ugly and unhelpful.
Each of the tests fail because our "gdb" test fixture tries to load
scylla-gdb.py and fails when the symbols it needs (e.g., "size_t")
cannot be found. So in this patch, we check once for the existance
of this symbol - and if missing we exit pytest instead of failing each
individual test.
Moreover, if loading scylla-gdb.py fails for some other unexpected
reason, let's exit the test as well, instead of failing each individual
test.
Fixes#10863.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#10937
Closes#10930
* github.com:scylladb/scylla:
test: perf_row_cache_update: Flush std output after each line
test: perf_row_cache_update: Drain background cleaner before starting the test
test: perf_row_cache_update: Measure memtable filling time
test: perf_row_cache_update: Respect preemption when applying mutations
test: perf_row_cache_update: Drop unused pk variable
Before this patch, the test cql-pytest/test_tools.py left behind
a temporary file in /tmp. It used pytest's "tmp_path_factory" feature,
but it doesn't remove temporary files it creates.
This patch removes the temporary file when the fixture using it ends,
but moreover, it puts the temporary file not in /tmp but rather next
to Scylla's data directory. That directory will be eventually removed
entirely, so even if we accidentally leave a file there, it will
eventually be deleted.
Fixes#10924
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closes#10929
There is a bug introduced in e74c3c8 (4.6.0) which makes memtable
reader skip one a range tombstone for a certain pattern of deletions
and under certain sequence of events.
_rt_stream contains the result of deoverlapping range tombstones which
had the same position, which were sipped from all the versions. The
result of deoverlapping may produce a range tombstone which starts
later, at the same position as a more recent tombstone which has not
been sipped from the partition version yet. If we consume the old
range tombstone from _rt_stream and then refresh the iterators, the
refresh will skip over the newer tombstone.
The fix is to drop the logic which drains _rt_stream so that
_rt_stream is always merged with partition versions.
For the problem to trigger, there have to be multiple MVCC versions
(at least 2) which contain deletions of the following form:
[a, c] @ t0
[a, b) @ t1, [b, d] @ t2
c > b
The proper sequence for such versions is (assuming d > c):
[a, b) @ t1,
[b, d] @ t2
Due to the bug, the reader will produce:
[a, b) @ t1,
[b, c] @ t0
The reader also needs to be preempted right before processing [b, d] @
t2 and iterators need to get invalidated so that
lsa_partition_reader::do_refresh_state() is called and it skips over
[b, d] @ t2. Otherwise, the reader will emit [b, d] @ t2 later. If it
does emit the proper range tombstone, it's possible that it will violate
fragment order in the stream if _rt_stream accumulated remainders
(possible with 3 MVCC versions).
The problem goes away once MVCC versions merge.
Fixes#10913Fixes#10830Closes#10914
