fe7e63f109
This patch series: - Increases the number of allowed scheduling groups to allow creation of `sl:driver` - Implements `create_driver_service_level` that creates `sl:driver` with shares=200 if it wasn't already created - Implements creation of `sl:driver` for new systems and tests in `raft_initialize_discovery_leader` - Modifies `topology_coordinator` to use create `sl:driver` after upgrades. - Implements using `sl:driver` for new connections in `transport/server` - Adds to `transport/server` recognition of driver's control connections and forcing them to keep using `sl:driver`. - Adds tests to verify the new functionality - Modifies existing tests to let them pass after `sl:driver` is added - Modifies the documentation to contain new `sl:driver` The changes were evaluated by a test with the following scenario ([test_connections-sl-driver.py](https://github.com/user-attachments/files/22021273/test_connections-sl-driver.py)): - Start ScyllaDB with one node - Create 1000 keyspaces, 1 table in each keyspace - Start `cassandra-stress` (`-rate threads=50 -mode native cql3`) - Run connection storm with 1000 session (100 python processes, 10 sessions each) The maximum latency during connection storm dropped **from 224.94ms to 41.43ms** (those numbers are average from 20 test executions, were max latency was in [140ms, 361ms] before change and [31.4ms, 61.5ms] after). The snippet of cassandra-stress output from the moment of connection storm: Before: ``` type total ops, op/s, pk/s, row/s, mean, med, .95, .99, .999, max, time, stderr, errors, gc: #, max ms, sum ms, sdv ms, mb ... total, 789206, 85887, 85887, 85887, 0.6, 0.3, 2.0, 2.0, 2.5, 5.0, 9.0, 0.09679, 0, 0, 0, 0, 0, 0 total, 909322, 120116, 120116, 120116, 0.4, 0.2, 1.9, 2.0, 2.1, 3.1, 10.0, 0.09053, 0, 0, 0, 0, 0, 0 total, 964392, 55070, 55070, 55070, 0.9, 0.4, 2.0, 4.5, 7.7, 18.9, 11.0, 0.09203, 0, 0, 0, 0, 0, 0 total, 975705, 11313, 11313, 11313, 4.4, 3.5, 6.5, 24.5, 82.7, 83.0, 12.0, 0.11713, 0, 0, 0, 0, 0, 0 total, 987548, 11843, 11843, 11843, 4.2, 3.5, 6.5, 33.7, 48.6, 51.5, 13.0, 0.13366, 0, 0, 0, 0, 0, 0 total, 995422, 7874, 7874, 7874, 6.3, 4.0, 7.7, 85.6, 112.9, 113.5, 14.0, 0.14753, 0, 0, 0, 0, 0, 0 total, 1007228, 11806, 11806, 11806, 4.3, 3.5, 6.5, 29.1, 43.8, 87.1, 15.0, 0.15598, 0, 0, 0, 0, 0, 0 total, 1012840, 5612, 5612, 5612, 8.2, 5.0, 11.5, 121.8, 166.6, 170.1, 16.0, 0.16535, 0, 0, 0, 0, 0, 0 total, 1016186, 3346, 3346, 3346, 13.4, 7.4, 20.1, 204.9, 207.6, 210.4, 17.0, 0.17405, 0, 0, 0, 0, 0, 0 total, 1025462, 9276, 9276, 9276, 6.3, 3.9, 9.6, 74.6, 206.8, 210.0, 18.0, 0.17800, 0, 0, 0, 0, 0, 0 total, 1035979, 10517, 10517, 10517, 4.8, 3.5, 6.7, 38.5, 82.6, 83.0, 19.0, 0.18120, 0, 0, 0, 0, 0, 0 total, 1047488, 11509, 11509, 11509, 4.3, 3.5, 6.0, 32.6, 72.3, 74.0, 20.0, 0.18334, 0, 0, 0, 0, 0, 0 total, 1077456, 29968, 29968, 29968, 1.7, 1.6, 2.9, 3.6, 7.0, 8.2, 21.0, 0.17943, 0, 0, 0, 0, 0, 0 total, 1105490, 28034, 28034, 28034, 1.8, 1.8, 3.5, 4.6, 5.3, 13.8, 22.0, 0.17609, 0, 0, 0, 0, 0, 0 total, 1132221, 26731, 26731, 26731, 1.9, 1.8, 3.8, 5.2, 8.4, 11.1, 23.0, 0.17314, 0, 0, 0, 0, 0, 0 total, 1162149, 29928, 29928, 29928, 1.7, 1.7, 3.0, 4.5, 8.0, 9.1, 24.0, 0.16950, 0, 0, 0, 0, 0, 0 ... ``` After: ``` type total ops, op/s, pk/s, row/s, mean, med, .95, .99, .999, max, time, stderr, errors, gc: #, max ms, sum ms, sdv ms, mb ... total, 822863, 94379, 94379, 94379, 0.5, 0.3, 2.0, 2.0, 2.1, 3.7, 9.0, 0.06669, 0, 0, 0, 0, 0, 0 total, 937337, 114474, 114474, 114474, 0.4, 0.2, 2.0, 2.0, 2.1, 3.4, 10.0, 0.06301, 0, 0, 0, 0, 0, 0 total, 986630, 49293, 49293, 49293, 1.0, 1.0, 2.0, 2.1, 17.9, 19.0, 11.0, 0.07318, 0, 0, 0, 0, 0, 0 total, 1026734, 40104, 40104, 40104, 1.2, 1.0, 2.0, 2.2, 6.3, 7.1, 12.0, 0.08410, 0, 0, 0, 0, 0, 0 total, 1066124, 39390, 39390, 39390, 1.3, 1.0, 2.0, 2.2, 2.6, 3.4, 13.0, 0.09108, 0, 0, 0, 0, 0, 0 total, 1103082, 36958, 36958, 36958, 1.3, 1.1, 2.1, 2.5, 3.1, 4.2, 14.0, 0.09643, 0, 0, 0, 0, 0, 0 total, 1141987, 38905, 38905, 38905, 1.3, 1.0, 2.0, 2.4, 11.4, 12.7, 15.0, 0.09894, 0, 0, 0, 0, 0, 0 total, 1180023, 38036, 38036, 38036, 1.3, 1.0, 2.0, 3.7, 5.6, 7.1, 16.0, 0.10070, 0, 0, 0, 0, 0, 0 total, 1216481, 36458, 36458, 36458, 1.4, 1.0, 2.1, 3.6, 4.7, 5.0, 17.0, 0.10210, 0, 0, 0, 0, 0, 0 total, 1256819, 40338, 40338, 40338, 1.2, 1.0, 2.0, 2.2, 3.5, 5.4, 18.0, 0.10173, 0, 0, 0, 0, 0, 0 total, 1295122, 38303, 38303, 38303, 1.3, 1.0, 2.0, 2.4, 21.0, 21.1, 19.0, 0.10136, 0, 0, 0, 0, 0, 0 total, 1334743, 39621, 39621, 39621, 1.3, 1.0, 2.0, 2.3, 3.3, 4.0, 20.0, 0.10055, 0, 0, 0, 0, 0, 0 total, 1375579, 40836, 40836, 40836, 1.2, 1.0, 2.0, 2.1, 3.4, 5.7, 21.0, 0.09927, 0, 0, 0, 0, 0, 0 total, 1415576, 39997, 39997, 39997, 1.2, 1.0, 2.0, 2.3, 3.2, 4.1, 22.0, 0.09807, 0, 0, 0, 0, 0, 0 total, 1449268, 33692, 33692, 33692, 1.5, 1.4, 2.5, 3.2, 4.2, 5.6, 23.0, 0.09800, 0, 0, 0, 0, 0, 0 total, 1471873, 22605, 22605, 22605, 2.2, 2.0, 4.8, 5.9, 7.0, 7.9, 24.0, 0.10015, 0, 0, 0, 0, 0, 0 ... ``` Fixes: https://github.com/scylladb/scylladb/issues/24411 This is a new feature, so no backport needed. Closes scylladb/scylladb#25412 * github.com:scylladb/scylladb: docs: workload-prioritization: add driver service level test: add test to verify use of `sl:driver` transport: use `sl:driver` to handle driver's control connections transport: whitespace only change in update_scheduling_group transport: call update_scheduling_group for non-auth connections generic_server: transport: start using `sl:driver` for new connections test: add test_desc_* for driver service level test: service_levels: add tests for sl:driver creation and removal test: add reload_raft_topology_state() to ScyllaRESTAPIClient service_level_controller: automatically create `sl:driver` service_level_controller: methods to create driver service level service_level_controller: handle special sl:driver in DESC output topology_coordinator: add service_level_controller reference system_keyspace: add service_level_driver_created test: add MAX_USER_SERVICE_LEVELS
Scylla unit tests using C++ and the Boost test framework
The source files in this directory are Scylla unit tests written in C++ using the Boost.Test framework. These unit tests come in three flavors:
-
Some simple tests that check stand-alone C++ functions or classes use Boost's
BOOST_AUTO_TEST_CASE. -
Some tests require Seastar features, and need to be declared with Seastar's extensions to Boost.Test, namely
SEASTAR_TEST_CASE. -
Even more elaborate tests require not just a functioning Seastar environment but also a complete (or partial) Scylla environment. Those tests use the
do_with_cql_env()ordo_with_cql_env_thread()function to set up a mostly-functioning environment behaving like a single-node Scylla, in which the test can run.
While we have many tests of the third flavor, writing new tests of this type should be reserved to white box tests - tests where it is necessary to inspect or control Scylla internals that do not have user-facing APIs such as CQL. In contrast, black-box tests - tests that can be written only using user-facing APIs, should be written in one of newer test frameworks that we offer - such as test/cqlpy or test/alternator (in Python, using the CQL or DynamoDB APIs respectively) or test/cql (using textual CQL commands), or - if more than one Scylla node is needed for a test - using the test/topology* framework.
Running tests
Because these are C++ tests, they need to be compiled before running.
To compile a single test executable row_cache_test, use a command like
ninja build/dev/test/boost/row_cache_test
You can also use ninja dev-test to build all C++ tests, or use
ninja deb-build to build the C++ tests and also the full Scylla executable
(however, note that full Scylla executable isn't needed to run Boost tests).
Replace "dev" by "debug" or "release" in the examples above and below to use the "debug" build mode (which, importantly, compiles the test with ASAN and UBSAN enabling on and helps catch difficult-to-catch use-after-free bugs) or the "release" build mode (optimized for run speed).
To run an entire test file row_cache_test, including all its test
functions, use a command like:
build/dev/test/boost/row_cache_test -- -c1 -m1G
to run a single test function test_reproduce_18045() from the longer test
file, use a command like:
build/dev/test/boost/row_cache_test -t test_reproduce_18045 -- -c1 -m1G
In these command lines, the parameters before the -- are passed to
Boost.Test, while the parameters after the -- are passed to the test code,
and in particular to Seastar. In this example Seastar is asked to run on one
CPU (-c1) and use 1G of memory (-m1G) instead of hogging the entire
machine. The Boost.Test option -t test_reproduce_18045 asks it to run just
this one test function instead of all the test functions in the executable.
Unfortunately, interrupting a running test with control-C while doesn't
work. This is a known bug (#5696). Kill a test with SIGKILL (-9) if you
need to kill it while it's running.
Boost tests can also be run using test.py - which is a script that provides a uniform way to run all tests in scylladb.git - C++ tests, Python tests, etc.
Execution with pytest
To run all tests with pytest execute
pytest test/boost
To execute all tests in one file, provide the path to the source filename as a parameter
pytest test/boost/aggregate_fcts_test.cc
Since it's a normal path, autocompletion works in the terminal out of the box.
To execute only one test function, provide the path to the source file and function name
pytest --mode dev test/boost/aggregate_fcts_test.cc::test_aggregate_avg
To provide a specific mode, use the next parameter --mode dev,
if parameter isn't provided pytest tries to use ninja mode_list to find out the compiled modes.
Parallel execution is controlled by pytest-xdist and the parameter -n auto.
This command starts tests with the number of workers equal to CPU cores.
The useful command to discover the tests in the file or directory is
pytest --collect-only -q --mode dev test/boost/aggregate_fcts_test.cc
That will return all test functions in the file.
To execute only one function from the test, you can invoke the output from the previous command.
However, suffix for mode should be skipped.
For example,
output shows in the terminal something like this test/boost/aggregate_fcts_test.cc::test_aggregate_avg.dev.
So to execute this specific test function, please use the next command
pytest --mode dev test/boost/aggregate_fcts_test.cc::test_aggregate_avg
Writing tests
Because of the large build time and build size of each separate test executable, it is recommended to put test functions into relatively large source files. But not too large - to keep compilation time of a single source file (during development) at reasonable levels.
When adding new source files in test/boost, don't forget to list the new source file in configure.py and also in CMakeLists.txt. The former is needed by our CI, but the latter is preferred by some developers.