in this change, following query timeouts config options are marked
live update-able:
- range_request_timeout_in_ms
- read_request_timeout_in_ms
- counter_write_request_timeout_in_ms
- cas_contention_timeout_in_ms
- truncate_request_timeout_in_ms
- write_request_timeout_in_ms
- request_timeout_in_ms
as per https://github.com/scylladb/scylladb/issues/10172,
> Many users would like to set the driver timers based on server timers.
> For example: expire a read timeout before or after the server read time
> out.
with this change, these options are *marked* live-updateable, but since
they are cached by their consumers locally, so we will have another commit
to update the local copies when these options get updated.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this function returns a const reference to member variable, so we
can mark it with the `const` specifier for better readability.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
`timeout_config` is not used by auth/common.hh. presumably, this
class is not a public interface exposed by auth, as it is not
inherently related auth. timeout_config is a shared setting across
related services, specifically, redis_server, thrift and cql_server.
so, in this change, let's drop this forward declaration.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
now that `redis_server::connection::timeout_config()` and
`redis_server::timeout_config()` are used nowhere, let's drop them.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this change should keep the timeout settings of handler_factory sync'ed
with the ones used by `thrift_server`. so far, the `timeout_config`
instance in `thrift_server` is not live-updateable, but in a follow-up
change, we will make it so. so, this change prepares the handler_factory
for a live-updateable timeout_config.
instead keeping a snapshot of the timeout_config, keep a reference of
it in handler_factory. the reference points to `thrift_server::_config`.
so despite that `thrift_server::_handler_factory` is a shared_ptr,
the member variable won't outlive its container, as the only reason to
have it as a shared_ptr is to appease the ctor of
`CassandraAsyncProcessorFactory`. and the constructed
`_processor_factory` is also a member variable of `thrift_server`, so we
won't take the risk of a dangling reference held by `handler_factory`.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
instead of copying the `config` parameter, move away from it.
this change also prepares for a non-copyable config. if the class
of `config` is not copyable, we will not be able to initialize
the member variable by copying from the given `config` parameter.
after the live-updateable config change, the `_config` member
variable will contain instances of utils::observer<>, which is
not copyable, but is move-constructable, hence in this change,
we just move away from the give `config`.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
* pass config via sharded_parameter
* initialize config using designated initializer
this change paves the road to servers with live-updateable timeout
options.
before this change, the servers initialize a domain specific combo
config, like `redis_server_config`, with the same instance of a
timeout_config, and pass the combox config as a ctor parameter to
construct each sharded service instance. but this design assumes
the value semantic of the config class, say, it should be copyable.
but if we want to use utils::updateable_value<> to get updated
option values, we would have to postpone the instantiation of the
config until the sharded service is about to be initialized.
so, in this change, instead of taking a domain specific config created
before hand, all services constructed with a `timeout_config` will
take a `sharded_parameter()` for creating the config. also, take
this opportunity to initialize the config using designated initializer.
for two reasons:
* less repeatings this way. we don't have to repeat the variable
name of the config being initialized for each member variable.
* prepare for some member variables which do not have a default
constructor. this applies to the timeout_config's updater which
will not have a default constructor, as it should be initialized
by db::config and a reference to the timeout_config to be updated.
we will update the `timeout_config` side in a follow-up commit.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Fixes#13332
The tests user the discriminator "system" as prefix to assume keyspaces are marked
"internal" inside scylla. This is not true in enterprise universe (replicated key
provider). It maybe/probably should, but that train is sailing right now.
Fix by removing one assert (not correct) and use actual API info in the alternator
test.
Closes#13333
this is the 15th changeset of a series which tries to give an overhaul to the CMake building system. this series has two goals:
- to enable developer to use CMake for building scylla. so they can use tools (CLion for instance) with CMake integration for better developer experience
- to enable us to tweak the dependencies in a simpler way. a well-defined cross module / subsystem dependency is a prerequisite for building this project with the C++20 modules.
this changeset includes following changes:
- build: cmake: add two missing tests
- build: cmake: port more cxxflags from configure.py
Closes#13262
* github.com:scylladb/scylladb:
build: cmake: add missing source files to idl and service
build: cmake: port more cxxflags from configure.py
build: cmake: add two missing tests
The concept is needed by enterprise functionality, but in the hunt for globals this sticks out and should be removed.
This is also partially prompted by the need to handle the keyspaces in the above set special on shutdown as well as startup. I.e. we need to ensure all user keyspaces are flushed/closed earlier then these. I.e. treat as "system" keyspace for this purpose.
These changes adds a "extension internal" keyspace set instead, which for now (until enterprise branches are updated) also included the "load_prio" set. However, it changes distributed loader to use the extension API interface instead, as well as adds shutdown special treatment to replica::database.
Closes#13335
* github.com:scylladb/scylladb:
datasbase: Flush/close "extension internal" keyspaces after other user ks
distributed_loader: Use extensions set of "extension internal" keyspaces
db::extentions: Add "extensions internal" keyspace set
we've been seeing errors like
```
10:39:36 gdb-add-index: [Was there no debuginfo? Was there already an index?]
10:39:36 readelf: /jenkins/workspace/scylla-master/next/scylla/build/dist/debug/redhat/BUILDROOT/scylla-5.3.0~dev-0.20230321.0f97d464d32b.x86_64/usr/lib/debug/opt/scylladb/libreloc/libc.so.6-5.3.0~dev-0.20230321.0f97d464d32b.x86_64.debug: Error: Unable to find program interpreter name
```
when strip.sh is processing *.debug elf images. this is caused by a
known issue, see
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1012107 . and this
error is not fatal. but it is very distracting when we are trying to
find errors in jenkins logging messages.
so, in this change, the stderr output from readelf is muted for higher
signal-noise ratio in the build logging message.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#13267
This is used for readiness API: /storage_service/rpc_server and the fix prevents from returning 'true' prematurely.
Some improvement for readiness was added in a51529dd15 but thrift implementation wasn't fully done.
Fixes https://github.com/scylladb/scylladb/issues/12376Closes#13319
* github.com:scylladb/scylladb:
thrift: return address in listen_addresses() only after server is ready
thrift: simplify do_start_server() with seastar:async
this is a part of a series migrating from `operator<<(ostream&, ..)` based formatting to fmtlib based formatting. the goal here is to enable fmtlib to print `bytes` and `gms::inet_address` without using ostream<<. also, this change removes all existing callers of `operator<<(ostream, const bytes &)` and `operator<<(ostream, const gms::inet_address&)`.
`gms::inet_address` related changes are included here in hope to demonstrate the usage of delimiter specifier of `fmt_hex` 's formatter.
Refs #13245Closes#13275
* github.com:scylladb/scylladb:
gms/inet_address: implement operator<< using fmt::formatter
treewide: use fmtlib to format gms::inet_address
gms/inet_address: specialize fmt::formatter<gms::inet_address>
bytes: implement formatting helpers using formatter
bytes: specialize fmt::formatter<bytes>
bytes: specialize fmt::formatter<fmt_hex>
bytes: mark fmt_hex::v `const`
Files with task manager repair module and related classes
are modified to be consistent with task manager compaction module.
Closes#13231
* github.com:scylladb/scylladb:
repair: rename repair_module
repair: add repair namespace to repair/task_manager_module.hh
repair: rename repair_task.hh
Refs #13334
Effectively treats keyspaces listed in "extension internal" as system keyspaces
w.r.t. shutdown/drain. This ensures all user keyspaces are fully flushed before
we disable these "internal" ones.
Refs #13334
To be populated early by extensions. Such a keyspace should be
1.) Started before user keyspaces
2.) Flushed/closed after user keyspaces
3.) For all other regards be considered "user".
* tools/cqlsh b9a606f...8769c4c (11):
> dist: redhat: provide only a single version
> pylib/setup, requirement.txt: remove Six
> setup: do not support python2
> install.sh: install files with correct permission in struct umask settings
> Remove unneed LC_ALL=en_US.UTF-8
> Support using other driver (datastax or older scylla ones)
> Fix RPM based downgrade command on scylla-cqlsh
> gitignore: ignore pylib/cqlshlib/__pycache__
> dist/redhat: add a proper changelog entry
> github actions: enable starting on tags
> Add support for building docker image
the goal of this change is to reduce the dependency on
`operator<<(ostream&, const gms::inet_address&)`.
this is not an exhaustive search-and-replace change, as in some
caller sites we have other dependencies to yet-converted ostream
printer, we cannot fix them all, this change only updates some
caller of `operator<<(ostream&, const gms::inet_address&)`.
Refs #13245
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this is a part of a series to migrating from `operator<<(ostream&, ..)`
based formatting to fmtlib based formatting. the goal here is to enable
fmtlib to print `gms::inet_address` with the help of fmt::ostream.
please note, the ':' delimiter is specified when printing the IPv6 address.
Refs #13245
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
some of these helpers prints a byte array using `to_hex()`, which
materializes a string instance and then drop it on the floor after
printing it to the given ostream. this hurts the performance, so
`fmt::print()` should be more performant in comparison to the
implementations based on `to_hex()`.
Refs #13245
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this is a part of a series to migrating from `operator<<(ostream&, ..)`
based formatting to fmtlib based formatting. the goal here is to enable
fmtlib to print bytes with the help of fmt::ostream.
Refs #13245
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this is a part of a series to migrating from `operator<<(ostream&, ..)`
based formatting to fmtlib based formatting. the goal here is to enable
fmtlib to print bytes_view with the help of fmt::ostream. because fmtlib
has its own specialization for fmt::formatter<std::basic_string_view<T>>,
we cannot just create a full specialization for std::basic_string_view<int8_t>,
otherwise fmtlib would complain that
> Mixing character types is disallowed.
so we workaround this using a delegate of fmt_hex.
Refs #13245
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
The `wait_for_normal_state_handled_on_boot` function waits until
`handle_state_normal` finishes for the given set of nodes. It was used
in `run_bootstrap_ops` and `run_replace_ops` to wait until NORMAL states
of existing nodes in the cluster are processed by the joining node
before continuing the joining process. One reason to do it is because at
the end of `handle_state_normal` the joining node might drop connections
to the NORMAL nodes in order to reestablish new connections using
correct encryption settings. In tests we observed that the connection
drop was happening in the middle of repair/streaming, causing
repair/streaming to abort.
Unfortunately, calling `wait_for_normal_state_handled_on_boot` in
`run_bootstrap_ops`/`run_replace_ops` is too late to fix all problems.
Before either of these two functions, we create a new CDC generation and
write the data to `system_distributed_everywhere.cdc_generation_descriptions_v2`.
In tests, the connections were sometimes dropped while this write was
in-flight. This would cause the write to never arrive to other nodes,
and the joining node would timeout waiting for confirmations.
To fix this, call `wait_for_normal_state_handled_on_boot` earlier in the
boot procedure, before `make_new_generation` call which does the write.
Fixes: #13302Closes#13317
* github.com:scylladb/scylladb:
storage_service: wait for normal state handlers earlier in the boot procedure
storage_service: bootstrap: wait for normal tokens to arrive in all cases
storage_service: extract get_nodes_to_sync_with helper
storage_service: return unordered_set from get_ignore_dead_nodes_for_replace
We need this so that we can have multi-partition mutations which are applied atomically. If they live on different shards, we can't guarantee atomic write to the commitlog.
Fixes: #12642Closes#13134
* github.com:scylladb/scylladb:
test_raft_upgrade: add a test for schema commit log feature
scylla_cluster.py: add start flag to server_add
ServerInfo: drop host_id
scylla_cluster.py: add config to server_add
scylla_cluster.py: add expected_error to server_start
scylla_cluster.py: ScyllaServer.start, refactor error reporting
scylla_cluster.py: fix ScyllaServer.start, reset cmd if start failed
raft: check if schema commitlog is initialized Refuse to boot if neither the schema commitlog feature nor force_schema_commit_log is set. For the upgrade procedure the user should wait until the schema commitlog feature is enabled before enabling consistent_cluster_management.
raft: move raft initialization after init_system_keyspace
database: rename before_schema_keyspace_init->maybe_init_schema_commitlog
raft: use schema commitlog for raft tables
init_system_keyspace: refactoring towards explicit load phases
listen_addresses() checks if _server variable is empty and after this
patch we assign (move) the value only after server is ready.
This is used for readiness API: /storage_service/rpc_server and the fix
prevents from returning 'true' prematurely. Some improvement for readiness
was added in a51529dd15 but thrift implementation
wasn't fully done.
Fixes#12376
* tools/python3 279b6c1...d2f57dd (3):
> dist: redhat: provide only a single version
> SCYLLA-VERSION-GEN: use -gt when comparing values
> SCYLLA-VERSION-GEN: remove unnecessary bashism
Task manager task implementations of classes that cover
cleanup keyspace compaction which can be started through
/storage_service/keyspace_compaction/ api.
Top level task covers the whole compaction and creates child
tasks on each shard.
Closes#12712
* github.com:scylladb/scylladb:
test: extend test_compaction_task.py to test cleanup compaction
compaction: create task manager's task for cleanup keyspace compaction on one shard
compaction: create task manager's task for cleanup keyspace compaction
api: add get_table_ids to get table ids from table infos
compaction: create cleanup_compaction_task_impl
as fmt_hex is a helper class for formatting the underlying `bytes_view`,
it does not mutate it, so mark the member variable const and mark
the parameter in its constructor const. this change also helps us to
use fmt_hex in the use case where the const semantics is expected.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
this is a part of a series migrating from `operator<<(ostream&, ..)` based formatting to fmtlib based formatting. the goal here is to enable fmtlib to print `range_tombstone` and `range_tombstone_change` without using ostream<<. also, this change removes all existing callers of `operator<<(ostream, const range_tombstone &)` and `operator<<(ostream, const range_tombstone_change &)`, and then removes these two `operator<<`s.
Refs #13245Closes#13260
* github.com:scylladb/scylladb:
mutation: drop operator<<(ostream, const range_tombstone{_change,} &)
mutation: use fmtlib to print range_stombstone{_change,}
mutation: mutation_fragment_v2: specialize fmt::formatter<range_tombstone_change>
mutation: range_tombstone: specialize fmt::formatter<range_tombstone>
at least, we need to access the declarations of exceptions, like`not_a_leader` and `dropped_entry`, so, instead of relying on other header to do this job for us, we should include the header which include the declaration. so, in this chance "raft.h" is include explicitly. also, include boost headers using "<path/to/header>` instead of "path/to/header` for more consistency.
Closes#13326
* github.com:scylladb/scylladb:
raft: include boost header using <path/to/header> not "path/to/header"
raft: include used header
* scripts/create-relocatable-package.py: add a command to print out
executables under libexec
* dist/debian/debian_files_gen.py: call create-relocatable-package.py
for a list of files under libexec and create source/include-binaries
with the list.
we repackage the precompiled binaries in the relocatable package into a debian source package using `./scylla/install.sh`, which edits the executable to use the specified dynamic library loader. but dpkg-source does not like this, as it wants to ensure that the files in original tarball (*.orig.tar.gz) is identical to the files in the source package created by dpkg-source.
so we have following failure when running reloc/build_deb.sh
```
dpkg-source: error: cannot represent change to scylla/libexec/scylla: binary file contents changed
dpkg-source: error: add scylla/libexec/scylla in debian/source/include-binaries if you want to store the modified binary in the debian tarball
dpkg-source: error: unrepresentable changes to source
dpkg-buildpackage: error: dpkg-source -b . subprocess returned exit status 1
debuild: fatal error at line 1182:
dpkg-buildpackage -rfakeroot -us -uc -ui failed
```
in this change, to address the build failure, as proposed by dpkg, the
path to the patched/edited executable is added to
`debian/source/include-binaries`. see the "Building" section in https://manpages.debian.org/bullseye/dpkg-dev/dpkg-source.1.en.html for more details.
please search `adjust_bin()` in `scylladb/install.sh` for more details.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#12722
this series applies some random cleanups to bloom_filter. these cleanups were the side products when the author was working on #13314 .
Closes#13315
* github.com:scylladb/scylladb:
bloom_filter: mark internal help function static
bloom_filter: add more constness to false positive rate tables
bloom_filter: use vector::back() when appropriate
before this change, we use `round(random.random(), 5)` for
the value of `bloom_filter_fp_chance` config option. there are
chances that this expression could return a number lower or equal
to 6.71e-05.
but we do have a minimal for this option, which is defined by
`utils::bloom_calculations::probs`. and the minimal false positive
rate is 6.71e-05.
we are observing test failures where the we are using 0 for
the option, and scylla right rejected it with the error message of
```
bloom_filter_fp_chance must be larger than 6.71e-05 and less than or equal to 1.0 (got 0)
```.
so, in this change, to address the test failure, we always use a number
slightly greater or equal to a number slightly greater to the minimum to
ensure that the randomly picked number is in the range of supported
false positive rate.
Fixes#13313
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#13314
When creating the reader, the lifecycle policy might return one that was saved on the last page and survived in the cache. This reader might have skipped some fast-forwarding ranges while sitting in the cache. To avoid using a reader reading a stale range (from the read's POV), check its read range and fast forward it if necessary.
Fixes: https://github.com/scylladb/scylladb/issues/12916Closes#12932
* github.com:scylladb/scylladb:
readers/multishard: shard_reader: fast-forward created reader to current range
readers/multishard: reader_lifecycle_policy: add get_read_range()
test/boost/multishard_mutation_query_test: paging: handle range becoming wrapping
The Wasm compilation is a slow, low priority task, so it should
not compete with reactor threads or the networking core.
To achieve that, we increase the niceness of the thread by 10.
An alternative solution would be to set the priority using
pthread_setschedparam, but it's not currently feasible,
because as long as we're using the SCHED_OTHER policy for our
threads, we cannot select any other priority than 0.
Closes#13307
Fixes https://github.com/scylladb/scylladb/issues/13106
This commit removes the information that BYPASS CACHE
is an Enterprise-only feature and replaces that info
with the link to the BYPASS CACHE description.
Closes#13316
<iterator> was introduced back in
1cf02cb9d8, but lexicographical_compare.hh
was extracted out in bdfc0aa748, since we
don't have any users of <iterator> in types.hh anymore, let's remove it.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#13327
s/%{version}/%{version}-%{release}/ in `Requires:` sections.
this enforces the runtime dependencies of exactly the same releases between scylla packages.
Fixes#13222
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#13229
* github.com:scylladb/scylladb:
dist/redhat: split Requires section into multiple lines
dist/redhat: enforce dependency on %{release} also
