Every table and sstable set keeps track of the total file size
of contained sstables.
Due to a feature request, we also want to keep track of the hypothetical
file size if Data files were uncompressed, to add a metric that
shows the compression ratio of sstables.
We achieve this by replacing the relevant `uint_64 bytes_on_disk`
counters everywhere with a struct that contains both the actual
(post-compression) size and the hypothetical pre-compression size.
This patch isn't supposed to change any observable behavior.
In the next patch, we will use these changes to add a new metric.
The latter is recommended in seastar, and the former was left as
compatibility alias. Latest seastar explicitly marks it as deprecated so
once the submodule is updated, compilation logs will explode.
Most of the patch is generated with
for f in $(git grep -l '\<distributed<[A-Za-z0-9:_]*>') ; do sed -e 's/\<distributed<\([A-Za-z0-9:_]*\)>/sharded<\1>/g' -i $f; done
for f in $(git grep -l distributed.hh); do sed -e 's/distributed.hh/sharded.hh/' -i $f ; done
and a small manual change in test/perf/perf.hh
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Closesscylladb/scylladb#26136
Prefer for specializing the local replication strategy,
local effective replication map, et. al byt defining
an is_local() predicate, similar to uses_tablets().
Note that is_vnode_based() still applies to local replication
strategy.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
This pull request is continuation of scylladb/scylladb#20688 - contents of the main commit are the same, the only change is the additional commit with a test.
Until this patch, the materialized view flow-control algorithm (https://www.scylladb.com/2018/12/04/worry-free-ingestion-flow-control/) used a constant delay_limit_us hard-coded to one second, which means that when the size of view-update backlog reached the maximum (10% of memory), we delay every request by an additional second - while smaller amounts of backlog will result in smaller delays.
This hard-coded one maximum second delay was considered *huge* - it will slow down a client with concurrency 1000 to just 1000 requests per second - but we already saw some workloads where it was not enough - such as a test workload running very slow reads at high concurrency on a slow machine, where a latency of over one second was expected for each read, so adding a one second latecy for writes wasn't having any noticable affect on slowing down the client.
So this patch replaces the hard-coded default with a live-updateable configuration parameter, `view_flow_control_delay_limit_in_ms`, which defaults to 1000ms as before.
Another useful way in which the new `view_flow_control_delay_limit_in_ms` can be used is to set it to 0. In that case, the view-update flow control always adds zero delay, and in effect - does absolutely nothing. This setting can be used in emergency situations where it is suspected that the MV flow control is not behaving properly, and the user wants to disable it.
The new parameter's help string mentions both these use cases of the parameter.
Fixes#18187
This is new functionality, no need to backport to any open source release.
Closesscylladb/scylladb#21647
* github.com:scylladb/scylladb:
materialized views: test for the MV delay configuration parameter
service: add injection for skipping view update backlog
materialized view: make flow-control maximum delay configurable
Information about view update backlog is propagated in two main ways:
- In RPCs that serve as responses to writes (MUTATION_DONE /
MUTATION_FAILED)
- Via gossip (application_state::VIEW_BACKLOG)
In tests, it can be benefical to disable the second mechanism. View
update backlog propagation via write responses happens synchronously
with respect to writes so it is easier to control and reason about,
while gossip is asynchronous and can overwrite the backlog that was
propagated via write responses.
Add `skip_updating_local_backlog_via_view_update_backlog_broker` error
injection which skips the logic that updates the local, per-endpoint
cache of view update backlogs from the gossip state.
In this rather large path we mode to address nodes in storage proxy by
host ids instead of ips. Some subsystems storage proxy calls to are
not yet converted to host ids, so we translate back and forth when we
interact with them.
Modernize the codebase by replacing Boost range adaptors with C++23 standard library views,
reducing external dependencies and leveraging modern C++ language features.
Key Changes:
- Replace `boost::adaptors::filtered` with `std::views::filter`
- Remove `#include <boost/range/adaptor/filtered.hpp>`
- Utilize standard library range views
Motivation:
- Reduce project's external dependency footprint
- Leverage standard library's range and view capabilities
- Improve long-term code maintainability
- Align with modern C++ best practices
Implementation Challenges and Considerations:
1. Range Conversion and Move Semantics
- `std::ranges::to` adaptor requires rvalue references
- Necessitated updates to variable and parameter constness
- Example: `cql3/restrictions/statement_restrictions.cc` modified to remove `const`
from `common` to enable efficient range conversion
2. Range Iteration and Mutation
- Range views may mutate internal state during iteration
- Cannot pass ranges by const reference in some scenarios
- Solution: Pass ranges by rvalue reference to explicitly indicate
state invalidation
Limitations:
- One instance of `boost::adaptors::filtered` temporarily preserved
due to lack of a C++23 alternative for `boost::join()`
- A comprehensive replacement will be addressed in a follow-up change
This change is part of our ongoing effort to modernize the codebase,
reducing external dependencies and adopting modern C++ practices.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closesscylladb/scylladb#21648
Currently, a view update backlog may reach an invalid state, when
its max is 0 and its relative_size() is NaN as a result. This can
be achieved either by constructing the backlog with a 0 max or by
modifying the max of an existing backlog. In particular, this
happens when creating the backlog using the default constructor.
In this patch the the default constructor is deleted and a check
is added to make sure that the max is different than 0 is added
to its constructor - if the check fails, we construct an empty
backlog instead, to handle the possibility of getting an invalid
backlog sent from a node with a version that's missing this check.
Additionally, we make the backlogs members private, exposing them
only through const getters.
On each shard of each node we store the view update backlogs of
other nodes to, depending on their size, delay responses to incoming
writes, lowering the load on these nodes and helping them get their
backlog to normal if it were too high.
These backlogs are propagated between nodes in two ways: the first
one is adding them to replica write responses. The seconds one
is gossiping any changes to the node's backlog every 1s. The gossip
becomes useful when writes stop to some node for some time and we
stop getting the backlog using the first method, but we still want
to be able to select a proper delay for new writes coming to this
node. It will also be needed for the mv admission control.
Currently, the backlog is gossiped from shard 0, as expected.
However, we also receive the backlog only on shard 0 and only
update this shard's backlogs for the other node. Instead, we'd
want to have the backlogs updated on all shards, allowing us
to use proper delays also when requests are received on shards
different than 0.
This patch changes the backlog update code, so that the backlogs
on all shards are updated instead. This will only be performed
up to once per second for each other node, and is done with
a lower priority, so it won't severly impact other work.
Fixes: scylladb/scylladb#19232Closesscylladb/scylladb#19268
Currently, there are 2 ways of sharing a backlog with other nodes: through
a gossip mechanism, and with responses to replica writes. In gossip, we
check each second if the backlog changed, and if it did we update other
nodes with it. However if the backlog for this node changed on another
node with a write response, the gossiped backlog is currently not updated,
so if after the response the backlog goes back to the value from the previous
gossip round, it will not get sent and the other node will stay with an
outdated backlog.
This patch changes this by notifying the gossip that a the backlog changed
since the last gossip round so a different backlog could have been send
through the response piggyback mechanism. With that information, gossip
will send an unchanged backlog to other nodes in the following gossip round.
Fixes: https://github.com/scylladb/scylladb/issues/18461
Currently, when a view update backlog is changed and sent
using gossip, we check whether the strtoll/strtoull
function used for reading the backlog returned
LLONG_MAX/ULLONG_MAX, signaling an error of a value
exceeding the type's limit, and if so, we do not store
it as the new value for the node.
However, the ULLONG_MAX value can also be used as the max
backlog size when sending empty backlogs that were never
updated. In theory, we could avoid sending the default
backlog because each node has its real backlog (based on
the node's memory, different than the ULLONG_MAX used in
the default backlog). In practice, if the node's
backlog changed to 0, the backlog sent by it will be
likely the default backlog, because when selecting
the biggest backlog across node's shards, we use the
operator<=>(), which treats the default backlog as
equal to an empty backlog and we may get the default
backlog during comparison if the backlog of some shard
was never changed (also it's the initial max value
we compare shard's backlogs against).
This patch removes the (U)LLONG_MAX check and replaces
it with the errno check, which is also set to ERANGE during
the strtoll error, and which won't prevent empty backlogs
from being read
Fixes: #18462Closesscylladb/scylladb#18560
The function `gms::version_generator::get_next_version()` can only be
called from shard 0 as it uses a global, unsynchronized counter to
issue versions. Notably, the function is used as a default argument for
the constructor of `gms::versioned_value` which is used from shorthand
constructors such as `versioned_value::cache_hitrates`,
`versioned_value::schema` etc.
The `cache_hitrate_calculator` service runs a periodic job which
updates the `CACHE_HITRATES` application state in the local gossiper
state. Each time the job is scheduled, it runs on the next shard (it
goes through shards in a round-robin fashion). The job uses the
`versioned_value::cache_hitrates` shorthand to create a
`versioned_value`, therefore risking a data race if it is not currently
executing on shard 0.
Fix the race by constructing the versioned value on shard 0.
get0() dates back from the days where Seastar futures carried tuples, and
get0() was a way to get the first (and usually only) element. Now
it's a distraction, and Seastar is likely to deprecate and remove it.
Replace with seastar::future::get(), which does the same thing.
Said method has a check on `_lb` not being null, before accessing it.
However, since 0e5754a, there was an unconditional access, adding an
entry for the local node. Move this inside the if, so it is covered by
the null-check. The only caller is the api (probably nodetool), the
worst that can happend is that they get completely empty load-map if
they call too early during startup.
Fixes: #16617Closesscylladb/scylladb#16659
Rather than calling on_change for each particular
application_state, pass an endpoint_state::map_type
with all changed states, to be processed as a batch.
In particular, thise allows storage_service::on_change
to update_peer_info once for all changed states.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
This series cleans up and hardens the endpoint locking design and
implementation in the gossiper and endpoint-state subscribers.
We make sure that all notifications (expect for `before_change`, that
apparently can be dropped) are called under lock_endpoint, as well as
all calls to gossiper::replicate, to serialize endpoint_state changes
across all shards.
An endpoint lock gets a unique permit_id that is passed to the
notifications and passed back by them if the notification functions call
the gossiper back for the same endpoint on paths that modify the
endpoint_state and may acquire the same endpoint lock - to prevent a
deadlock.
Fixes scylladb/scylladb#14838
Refs scylladb/scylladb#14471
Closes#14845
* github.com:scylladb/scylladb:
gossiper: replicate: ensure non-null permit
gossiper: add_saved_endpoint: lock_endpoint
gossiper: mark_as_shutdown: lock_endpoint
gossiper: real_mark_alive: lock_endpoint
gossiper: advertise_token_removed: lock_endpoint
gossiper: do_status_check: lock_endpoint
gossiper: remove_endpoint: lock_endpoint if needed
gossiper: force_remove_endpoint: lock_endpoint if needed
storage_service: lock_endpoint when removing node
gossiper: use permit_id to serialize state changes while preventing deadlocks
gossiper: lock_endpoint: add debug messages
utils: UUID: make default tagged_uuid ctor constexpr
gossiper: lock_endpoint must be called on shard 0
gossiper: replicate: simplify interface
gossiper: mark_as_shutdown: make private
gossiper: convict: make private
gossiper: mark_as_shutdown: do not call convict
Pass permit_id to subscribers when we acquire one
via lock_endpoint. The subscribers then pass it back to
gossiper for paths that acquire lock_endpoint for
the same endpoint, to detect nested locks when the endpoint
is locked with the same permit_id.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
As a preparation for ensuring access safety for column families
related maps, add tables_metadata, access to members of which
would be protected by rwlock.
and provide accessor functions to get them.
1. So they can't be modified by mistake, as the versioned value is
immutable. A new value must have a higher version.
2. Before making the version a strong gms::version_type.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
these warnings are found by Clang-17 after removing
`-Wno-unused-lambda-capture` and '-Wno-unused-variable' from
the list of disabled warnings in `configure.py`.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Define table_id as a distinct utils::tagged_uuid modeled after raft
tagged_id, so it can be differentiated from other uuid-class types,
in particular from table_schema_version.
Fixes#11207
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
This patch avoids unncessary CACHE_HITRATES updates through gossip.
After this patch:
Publish CACHE_HITRATES in case:
- We haven't published it at all
- The diff is bigger than 1% and we haven't published in the last 5 seconds
- The diff is really big 10%
Note: A peer node can know the cache hitrate through read_data
read_mutation_data and read_digest RPC verbs which have cache_temperature in
the response. So there is no need to update CACHE_HITRATES through gossip in
high frequency.
We do the recalculation faster if the diff is bigger than 0.01. It is useful to
do the calculation even if we do not publish the CACHE_HITRATES though gossip,
since the recalculation will call the table->set_global_cache_hit_rate to set
the hitrate.
Fixes#5971Closes#11079
After fcb8d040 ("treewide: use Software Package Data Exchange
(SPDX) license identifiers"), many dual-licensed files were
left with empty comments on top. Remove them to avoid visual
noise.
Closes#10562
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
Move replica-oriented classes to the replica namespace. The main
classes moved are ::database, ::keyspace, and ::table, but a few
ancillary classes are also moved. There are certainly classes that
should be moved but aren't (like distributed_loader) but we have
to start somewhere.
References are adjusted treewide. In many cases, it is obvious that
a call site should not access the replica (but the data_dictionary
instead), but that is left for separate work.
scylla-gdb.py is adjusted to look for both the new and old names.
The database, keyspace, and table classes represent the replica-only
part of the objects after which they are named. Reading from a table
doesn't give you the full data, just the replica's view, and it is not
consistent since reconciliation is applied on the coordinator.
As a first step in acknowledging this, move the related files to
a replica/ subdirectory.
recalculate_hitrates() is defined to return future<lowres_clock::duration>
but actually returns future<std::chrono::milliseconds>. This fails
if the two types are not the same.
Fix by returning the declared type.
The calculator needs to update its app-state on gossiper. Keeping
a reference is safe -- gossiper starts early, the calculator -- at
the very very end, stop in reverse.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Returning a function parameter guarantees copy elision and does not
require a std::move(). Enable -Wredundant-move to warn us that the
move is unneeded, and gain slightly more readable code. A few violations
are trivially adjusted.
Closes#9004
Eliminate not used includes and replace some more includes
with forward declarations where appropriate.
Signed-off-by: Pavel Solodovnikov <pa.solodovnikov@scylladb.com>
The database.hh is the central recursive-headers knot -- it has ~50
includes. This patch leaves only 34 (it remains the champion though).
Similar thing for database.cc.
Both changes help the latter compile ~4% faster :)
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Message-Id: <20210414183107.30374-1-xemul@scylladb.com>
The blacklog of current and max in VIEW_BACKLOG is not update but the
nodes are updating VIEW_BACKLOG all the time. For example:
```
INFO 2020-03-06 17:13:46,761 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.3, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486026590,718)
INFO 2020-03-06 17:13:46,821 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.2, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486026531,742)
INFO 2020-03-06 17:13:47,765 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.3, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486027590,721)
INFO 2020-03-06 17:13:47,825 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.2, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486027531,745)
INFO 2020-03-06 17:13:48,772 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.3, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486028590,726)
INFO 2020-03-06 17:13:48,833 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.2, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486028531,750)
INFO 2020-03-06 17:13:49,772 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.3, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486029590,729)
INFO 2020-03-06 17:13:49,832 [shard 0] storage_service - Update system.peers table: endpoint=127.0.0.2, app_state=VIEW_BACKLOG, versioned_value=Value(0:18446744073709551615:1583486029531,753)
```
The downside of such updates:
- Introduces more gossip exchange traffic
- Updates system.peers all the time
The extra unnecessary gossip traffic is fine to a cluster in a good
shape but when some of the nodes or shards are loaded, such messages and
the handling of such messages can make the system even busy.
With this patch, VIEW_BACKLOG is updated only when the backlog is really
updated.
Btw, we can even make the update only when the change of the backlog is
great than a threshold, e.g., 5%, which can reduce the traffic even
further.
Fixes#5970
The new guarantees are a bit better IMHO:
Once a subscriber is removed, it is never notified. This was not true
in the old code since it would iterate over a copy that would still
have that subscriber.
Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
load_broadcaster::stop_broadcasting uses shared_from_this(). Since
that is the only reference that the produced shared_ptr knows of, it
is deleted immediately. Fix that by also using a shared_ptr in
load_meter.
Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
