Currently, most of the failures that occur during CQL reads or writes are reported using C++ exceptions. Although the seastar framework avoids most of the cost of unwinding by keeping exceptions in futures as `std::exception_ptr`s, the exceptions need to be inspected at various points for the purposes of accounting metrics or converting them to a CQL error response. Analyzing the value and type of an exception held by `std::exception_ptr`'s cannot be done without rethrowing the exception, and that can be very costly even if the exception is immediately caught. Because of that, exceptions are not a good fit for reporting failures which happen frequently during overload, especially if the CPU is the bottleneck.
This PR introduces facilities for reporting exceptions as values using the boost::outcome library. As a first step, the need to use exceptions for reporting timeouts was eliminated for regular and batch writes, and no exceptions are thrown between creation of a `mutation_write_timeout_exception` and its serialization as a CQL response in the `cql_server`.
The types and helpers introduced here can be reused in order to migrate more exceptions and exception paths in a similar fashion.
Results of `perf_simple_query --smp 1 --operations-per-shard 1000000`:
Master (00a9326ae7)
128789.53 tps ( 82.2 allocs/op, 12.2 tasks/op, 49245 insns/op)
This PR
127072.93 tps ( 82.2 allocs/op, 12.2 tasks/op, 49356 insns/op)
The new version seems to be slower by about 100 insns/op, fortunately not by much (about 0.2%).
Tests: unit(dev), unit(result_utils_test, debug)
Closes#10014
* github.com:scylladb/scylla:
cql_test_env: optimize handling result_message::exception
transport/server: handle exceptions from coordinator_result without throwing
transport/server: propagate coordinator_result to the error handling code
transport/server: unwrap the exception result_message in process_xyz_internal
query_processor: add exception-returning variants of execute_ methods
modification_statement: propagate failed result through result_message::exception
batch_statement: propagate failed result through result_message::exception
cql_statement: add `execute_without_checking_exception_message`
result_message: add result_message::exception
storage_proxy: change mutate_with_triggers to return future<result<>>
storage_proxy: add mutate_atomically_result
storage_proxy: return result<> from mutate_result
storage_proxy: return result<> from mutate_internal
storage_proxy: properly propagate future from mutate_begin to mutate_end
storage_proxy: handle exceptions as values in mutate_end
storage_proxy: let mutate_end take a future<result<>>
storage_proxy: resultify mutate_begin
storage_proxy: use result in the _ready future of write handlers
storage_proxy: introduce helpers for dealing with results
exceptions: add coordinator_exception_container and coordinator_result
utils: add result utils
utils: add exception_container
In order to propagate exceptions as values through the CQL layer with
minimal modifications to the interfaces, a new result_message type is
introduced: result_message::exception. Similarly to
result_message::bounce_to_shard, this is an internal type which is
supposed to be handled before being returned to the client.
Seastar uses POSIX IO for output in addition to C++ iostreams,
e.g. in print_safe(), where it write()s directly to stdout.
Instead of manipulating C++ output streams to reset
stdout/log files, reopen the underlying file descriptors
to output/log files.
Fixes#9962 "cql_repl prints junk into the log"
Message-Id: <20220204205032.1313150-1-kostja@scylladb.com>
Instead of lengthy blurbs, switch to single-line, machine-readable
standardized (https://spdx.dev) license identifiers. The Linux kernel
switched long ago, so there is strong precedent.
Three cases are handled: AGPL-only, Apache-only, and dual licensed.
For the latter case, I chose (AGPL-3.0-or-later and Apache-2.0),
reasoning that our changes are extensive enough to apply our license.
The changes we applied mechanically with a script, except to
licenses/README.md.
Closes#9937
This allows us to forward-declare raw_selector, which in turn reduces
indirect inclusions of expression.hh from 147 to 58, reducing rebuilds
when anything in that area changes.
Includes that were lost due to the change are restored in individual
translation units.
Closes#9434
Timeout config is now stored in each connection, so there's no point
in tracking it inside each query as well. This patch removes
timeout_config from query_options and follows by removing now
unnecessary parameters of many functions and constructors.
libstdc++'s std::regex uses recursion[1], with a depth controlled by the
input. Together with clang's debug mode, this overflows the stack.
Use boost::regex instead, which is immune to the problem.
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86164Closes#7378
Previously system.paxos TTL was set as max(3h, gc_grace_seconds).
Introduce new per-table option named `paxos_grace_seconds` to set
the amount of seconds which are used to TTL data in paxos tables
when using LWT queries against the base table.
Default value is equal to `DEFAULT_GC_GRACE_SECONDS`,
which is 10 days.
This change allows to easily test various issues related to paxos TTL.
Fixes#6284
Tests: unit (dev, debug)
Co-authored-by: Alejo Sanchez <alejo.sanchez@scylladb.com>
Signed-off-by: Alejo Sanchez <alejo.sanchez@scylladb.com>
Signed-off-by: Pavel Solodovnikov <pa.solodovnikov@scylladb.com>
Message-Id: <20200816223935.919081-1-pa.solodovnikov@scylladb.com>
C++20 introduced `contains` member functions for maps and sets for
checking whether an element is present in the collection. Previously
`count` function was often used in various ways.
`contains` does not only express the intend of the code better but also
does it in more unified way.
This commit replaces all the occurences of the `count` with the
`contains`.
Tests: unit(dev)
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Message-Id: <b4ef3b4bc24f49abe04a2aba0ddd946009c9fcb2.1597314640.git.piotr@scylladb.com>
In order to eventually switch to a single JSON library,
most of the libjsoncpp usage is dropped in favor of rjson.
Unfortunately, one usage still remains:
test/utils/test_repl utility heavily depends on the *exact textual*
format of its output JSON files, so replacing a library results
in all tests failing because of differences in formatting.
It is possible to force rjson to print its documents in the exact
matching format, but that's left for later, since the issue is not
critical. It would be nice though if our test suite compared
JSON documents with a real JSON parser, since there are more
differences - e.g. libjsoncpp keeps children of the object
sorted, while rapidjson uses an unordered data structure.
This change should cause no change in semantics, it strives
just to replace all usage of libjsoncpp with rjson.
This removes the need to include reactor.hh, a source of compile
time bloat.
In some places, the call is qualified with seastar:: in order
to resolve ambiguities with a local name.
Includes are adjusted to make everything compile. We end up
having 14 translation units including reactor.hh, primarily for
deprecated things like reactor::at_exit().
Ref #1
Alternator and CDC, apart from enabling their experimental features,
need to have their schema extensions registered. This patch adds missing
registration of schema extensions to cql_repl, so that cql tests written
with Alternator or CDC in mind will properly work.
cql_repl is a simple program which reads CQL from stdin,
executes it, and writes results to stdout.
It support --input, --output and --log options.
--log is directed to cql_test.log by default.
--input is stdin by default
--output is stdout by default.
The result set output is print with a basic
JSON visitor.
