This PR introduces an experimental feature called "tablets". Tablets are
a way to distribute data in the cluster, which is an alternative to the
current vnode-based replication. Vnode-based replication strategy tries
to evenly distribute the global token space shared by all tables among
nodes and shards. With tablets, the aim is to start from a different
side. Divide resources of replica-shard into tablets, with a goal of
having a fixed target tablet size, and then assign those tablets to
serve fragments of tables (also called tablets). This will allow us to
balance the load in a more flexible manner, by moving individual tablets
around. Also, unlike with vnode ranges, tablet replicas live on a
particular shard on a given node, which will allow us to bind raft
groups to tablets. Those goals are not yet achieved with this PR, but it
lays the ground for this.
Things achieved in this PR:
- You can start a cluster and create a keyspace whose tables will use
tablet-based replication. This is done by setting `initial_tablets`
option:
```
CREATE KEYSPACE test WITH replication = {'class': 'NetworkTopologyStrategy',
'replication_factor': 3,
'initial_tablets': 8};
```
All tables created in such a keyspace will be tablet-based.
Tablet-based replication is a trait, not a separate replication
strategy. Tablets don't change the spirit of replication strategy, it
just alters the way in which data ownership is managed. In theory, we
could use it for other strategies as well like
EverywhereReplicationStrategy. Currently, only NetworkTopologyStrategy
is augmented to support tablets.
- You can create and drop tablet-based tables (no DDL language changes)
- DML / DQL work with tablet-based tables
Replicas for tablet-based tables are chosen from tablet metadata
instead of token metadata
Things which are not yet implemented:
- handling of views, indexes, CDC created on tablet-based tables
- sharding is done using the old method, it ignores the shard allocated in tablet metadata
- node operations (topology changes, repair, rebuild) are not handling tablet-based tables
- not integrated with compaction groups
- tablet allocator piggy-backs on tokens to choose replicas.
Eventually we want to allocate based on current load, not statically
Closes#13387
* github.com:scylladb/scylladb:
test: topology: Introduce test_tablets.py
raft: Introduce 'raft_server_force_snapshot' error injection
locator: network_topology_strategy: Support tablet replication
service: Introduce tablet_allocator
locator: Introduce tablet_aware_replication_strategy
locator: Extract maybe_remove_node_being_replaced()
dht: token_metadata: Introduce get_my_id()
migration_manager: Send tablet metadata as part of schema pull
storage_service: Load tablet metadata when reloading topology state
storage_service: Load tablet metadata on boot and from group0 changes
db, migration_manager: Notify about tablet metadata changes via migration_listener::on_update_tablet_metadata()
migration_notifier: Introduce before_drop_keyspace()
migration_manager: Make prepare_keyspace_drop_announcement() return a future<>
test: perf: Introduce perf-tablets
test: Introduce tablets_test
test: lib: Do not override table id in create_table()
utils, tablets: Introduce external_memory_usage()
db: tablets: Add printers
db: tablets: Add persistence layer
dht: Use last_token_of_compaction_group() in split_token_range_msb()
locator: Introduce tablet_metadata
dht: Introduce first_token()
dht: Introduce next_token()
storage_proxy: Improve trace-level logging
locator: token_metadata: Fix confusing comment on ring_range()
dht, storage_proxy: Abstract token space splitting
Revert "query_ranges_to_vnodes_generator: fix for exclusive boundaries"
db: Exclude keyspace with per-table replication in get_non_local_strategy_keyspaces_erms()
db: Introduce get_non_local_vnode_based_strategy_keyspaces()
service: storage_proxy: Avoid copying keyspace name in write handler
locator: Introduce per-table replication strategy
treewide: Use replication_strategy_ptr as a shorter name for abstract_replication_strategy::ptr_type
locator: Introduce effective_replication_map
locator: Rename effective_replication_map to vnode_effective_replication_map
locator: effective_replication_map: Abstract get_pending_endpoints()
db: Propagate feature_service to abstract_replication_strategy::validate_options()
db: config: Introduce experimental "TABLETS" feature
db: Log replication strategy for debugging purposes
db: Log full exception on error in do_parse_schema_tables()
db: keyspace: Remove non-const replication strategy getter
config: Reformat
In order not to copy the rvalue consumer arg -- instantly convert it
into value. No other tricks.
Indentation is deliberately left broken.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
There is a `query_mutations` function which loads the entire contents of
a given table into memory. There was no function for e.g. loading just a
single partition in the form of mutations. Introduce one.
Once a new CDC generation is committed to the cluster by the topology
coordinator, we also need to publish it to the user-facing description
tables so CDC applications know which streams to read from.
This uses regular distributed table writes underneath (tables living
in the `system_distributed` keyspace) so it requires `token_metadata`
to be nonempty. We need a hack for the case of bootstrapping the
first node in the cluster - turning the tokens into normal tokens
earlier in the procedure in `token_metadata`, but this is fine for the
single-node case since no streaming is happening.
The `CDC_GENERATIONS_V3` table schema is a copy-paste of the
`CDC_GENERATIONS_V2` schema. The difference is that V2 lives in
`system_distributed_keyspace` and writes to it are distributed using
regular `storage_proxy` replication mechanisms based on the token ring.
The V3 table lives in `system_keyspace` and any mutations written to it
will go through group 0.
Also extend the `TOPOLOGY` schema with new columns:
- `new_cdc_generation_data_uuid` will be stored as part of a bootstrapping
node's `ring_slice`, it stores UUID of a newly introduced CDC
generation which is used as partition key for the `CDC_GENERATIONS_V3`
table to access this new generation's data. It's a regular column,
meaning that every row (corresponding to a node) will have its own.
- `current_cdc_generation_uuid` and `current_cdc_generation_timestamp`
together form the ID of the newest CDC generation in the cluster.
(the uuid is the data key for `CDC_GENERATIONS_V3`, the timestamp is
when the CDC generation starts operating). Those are static columns
since there's a single newest CDC generation.
Add new columns to the `system.topology` table: `shard_count` and
`ignore_msb`. When a node bootstraps or restarts and observes that the
values stored in `topology` are different than the local values, it
updates them. This is done in the `update_topology_with_local_metadata`
function (the 'metadata' here being the two values).
Additional flag persisted in `system.scylla_local` is used to safely
avoid performing read barriers when the values didn't change on node
restart. A comment in `update_topology_with_local_metadata` explains why
this flag is needed.
An example use case where `shard_count` and `ignore_msb` are needed is
creating CDC generations.
Fixes: #13508
The schema is
CREATE TABLE system.sstables (
location text,
generation bigint,
format text,
status text,
uuid uuid,
version text,
PRIMARY KEY (location, generation)
)
A sample entry looks like:
location | generation | format | status | uuid | version
---------------------------------------------------------------------+------------+--------+--------+--------------------------------------+---------
/data/object_storage_ks/test_table-d096a1e0ad3811ed85b539b6b0998182 | 2 | big | sealed | d0a743b0-ad38-11ed-85b5-39b6b0998182 | me
The uuid field points to the "folder" on the storage where the sstable
components are. Like this:
s3
`- test_bucket
`- f7548f00-a64d-11ed-865a-0c1fbc116bb3
`- Data.db
- Index.db
- Filter.db
- ...
It's not very nice that the whole /var/lib/... path is in fact used as
location, it needs the PR #12707 to fix this place.
Also, the "status" part is not yet fully functional, it only supports
three options:
- creating -- the same as TemporaryTOC file exists on disk
- sealed -- default state
- deleting -- the analogy for the deletion log on disk
The latter needs support from the distributed_loader, which's not yet
there. In fact, distributes_loader also needs to be patched to actualy
select entries from this table on load. Also it needs the mentioned
PR #12707 to support staging and quarantine sstables.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The latter is the only user of the class. This keeps system keyspace
code free from unrelated logic and from raft::server_id type.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
In order to keep system keyspace free from group0 logic and from the
service::group0_upgrade_state type
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
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
There was an attempt to cut feature-service -> system-keyspace dependency (#13172) which turned out to require more changes. Here's a preparation squeezing from this future work.
This set
- leaves only batch-enabling API in feature service
- keeps the need for async context in feature service
- narrows down system keyspace features API to only load and store records
- relaxes features updating logic in sys.ks.
- cosmetic
Closes#13264
* github.com:scylladb/scylladb:
feature_service: Indentation fix after previous patch
feature_service: Move async context into enable()
system_keyspace: Refactor local features load/save helpers
feature_service: Mark supported_feature_set() const
feature_service: Remove single feature enabling method
boot: Enable features in batch
gossiper: Enable features in batch
We aim (#12642) to use the schema commit log
for raft tables. Now they are loaded at
the first call to init_system_keyspace in
main.cc, but the schema commitlog is only
initialized shortly before the second
call. This is important, since the schema
commitlog initialization
(database::before_schema_keyspace_init)
needs to access schema commitlog feature,
which is loaded from system.scylla_local
and therefore is only available after the
first init_system_keyspace call.
So the idea is to defer the loading of the raft tables
until the second call to init_system_keyspace,
just as it works for schema tables.
For this we need a tool to mark which tables
should be loaded in the first or second phase.
To do this, in this patch we introduce system_table_load_phase
enum. It's set in the schema_static_props for schema tables.
It replaces the system_keyspace::table_selector in the
signature of init_system_keyspace.
The call site for populate_keyspace in init_system_keyspace
was changed, table_selector.contains_keyspace was replaced with
db.local().has_keyspace. This check prevents calling
populate_keyspace(system_schema) on phase1, but allows for
populate_keyspace(system) on phase2 (to init raft tables).
On this second call some tables from system keyspace
(e.g. system.local) may have already been populated on phase1.
This check protects from double-populating them, since every
populated cf is marked as ready_for_writes.
Introduce load_local_enabled_features() and save_local_enabled_features()
that get and put std::set<sstring> with feature names (and perform set to
string and back conversions on their own). They look natural next to
existing sys.ks. methods to get/set local-supported features and peer
features.
Using the new API, the more generic functions to preserve individual
features and load them on startup can become much shorter and cleaner.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
This patch continues the refactoring, now we move
wait_for_sync_to_commitlog property from schema_builder to
schema_static_props.
The patch replaces schema_builder::set_wait_for_sync_to_commitlog
and is_extra_durable with two register_static_configurator,
one in system_keyspace and another in system_distributed_keyspace.
They correspond to the two parts of the original disjunction
in schema_tables::is_extra_durable.
Now when both callers of this method are non-static, it can be made
non-static too. While at it make two more changes:
1. move the thing to private
2. remove explicit cql3::query_processor::cache_internal::yes argument,
the system_keyspace::execute_cql() applies it on itw own
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Now the call is done via the system_keyspace instance, so it can be
unmarked static and can use the local query processor instead of global
qctx.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Schema related files are moved there. This excludes schema files that
also interact with mutations, because the mutation module depends on
the schema. Those files will have to go into a separate module.
Closes#12858
There's a bunch of them used by mainly view_builder and also by the API
and storage_service. All use global qctx to make its job, now when the
callers have main-local sharded<system_keysace> references they can be
made non-static.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
It's only called on cluster-join from storage_service which has the
local system_keyspace reference and it's already started by that time.
This allows removing few more occurrences of global qctx.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Add a new virtual table `system.raft_state` that shows the currently
operating Raft configuration for each present group. The schema is the
same as `system.raft_snapshot_config` (the latter shows the config from
the last snapshot). In the future we plan to add more columns to this
table, showing more information (like the current leader and term),
hence the generic name.
Adding the table requires some plumbing of
`sharded<raft_group_registry>&` through function parameters to make it
accessible from `register_virtual_tables`, but it's mostly
straightforward.
Also added some APIs to `raft_group_registry` to list all groups and
find a given group (returning `nullptr` if one isn't found, not throwing
an exception).
Make it clear that the table stores the snapshot configuration, which is
not necessarily the currently operating configuration (the last one
appended to the log).
In the future we plan to have a separate virtual table for showing the
currently operating configuration, perhaps we will call it
`system.raft_config`.
It is equal to (if present) HOST_ID and no longer used for anything.
The application state was only gossiped if `experimental-features`
contained `raft`, so we can free this slot.
Similarly, `raft_server_id`s were only persisted in `system.peers` if
the `SUPPORTS_RAFT` cluster feature was enabled, which happened only
when `experimental-features` contained `raft`. The `raft_server_id`
field in the schema was also introduced recently in `master` and didn't
get to be in a release yet. Given either of these reasons, we can remove
this field safely.
The Host ID now uniquely identifies a node (we no longer steal it during
node replace) and Raft is still experimental. We can reuse the Host ID
of a node as its Raft ID. This will allow us to remove and simplify a
lot of code.
With this we can already remove some dead code in this commit.
Now that the local host_id is never changed externally
(by the storage_service upon replace-node),
the method can be made private and be used only for initializing the
local host_id to a random one.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
1) make address map API flexible
Before this patch:
- having a mapping without an actual IP address was an
internal error
- not having a mapping for an IP address was an internal
error
- re-mapping to a new IP address wasn't allowed
After this patch:
- the address map may contain a mapping
without an actual IP address, and the caller must be prepared for it:
find() will return a nullopt. This happens when we first add an entry
to Raft configuration and only later learn its IP address, e.g. via
gossip.
- it is allowed to re-map an existing entry to a new address;
2) subscribe to gossip notifications
Learning IP addresses from gossip allows us to adjust
the address map whenever a node IP address changes.
Gossiper is also the only valid source of re-mapping, other sources
(RPC) should not re-map, since otherwise a packet from a removed
server can remap the id to a wrong address and impact liveness of a Raft
cluster.
3) prompt address map state with app state
Initialize the raft address map with initial
gossip application state, specifically IPs of members
of the cluster. With this, we no longer need to store
these IPs in Raft configuration (and update them when they change).
The obvious drawback of this approach is that a node
may join Raft config before it propagates its IP address
to the cluster via gossip - so the boot process has to
wait until it happens.
Gossip also doesn't tell us which IPs are members of Raft configuration,
so we subscribe to Group0 configuration changes to mark the
members of Raft config "non-expiring" in the address translation
map.
Thanks to the changes above, Raft configuration no longer
stores IP addresses.
We still keep the 'server_info' column in the raft_config system table,
in case we change our mind or decide to store something else in there.
There are two places to patch: .start() and .setup() and both only need
snitch to get local dc/rack from, nothing more. Thus both can live with
the explicit argument for now
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
"
There's an ongoing effort to move the endpoint -> {dc/rack} mappings
from snitch onto topology object and this set finalizes it. After it the
snitch service stops depending on gossiper and system keyspace and is
ready for de-globalization. As a nice side-effect the system keyspace no
longer needs to maintain the dc/rack info cache and its starting code gets
relaxed.
refs: #2737
refs: #2795
"
* 'br-snitch-dont-mess-with-topology-data-2' of https://github.com/xemul/scylla: (23 commits)
system_keyspace: Dont maintain dc/rack cache
system_keyspace: Indentation fix after previous patch
system_keyspace: Coroutinuze build_dc_rack_info()
topology: Move all post-configuration to topology::config
snitch: Start early
gossiper: Do not export system keyspace
snitch: Remove gossiper reference
snitch: Mark get_datacenter/_rack methods const
snitch: Drop some dead dependency knots
snitch, code: Make get_datacenter() report local dc only
snitch, code: Make get_rack() report local rack only
storage_service: Populate pending endpoint in on_alive()
code: Populate pending locations
topology: Put local dc/rack on topology early
topology: Add pending locations collection
topology: Make get_location() errors more verbose
token_metadata: Add config, spread everywhere
token_metadata: Hide token_metadata_impl copy constructor
gosspier: Remove messaging service getter
snitch: Get local address to gossip via config
...
In preparation for supporting IP address changes of Raft Group 0:
1) Always use start_server_for_group0() to start a server for group 0.
This will provide a single extension point when it's necessary to
prompt raft_address_map with gossip data.
2) Don't use raft::server_address in discovery, since going forward
discovery won't store raft::server_address. On the same token stop
using discovery::peer_set anywhere outside discovery (for persistence),
use a peer_list instead, which is easier to marshal.
Closes#11676
* github.com:scylladb/scylladb:
raft: (discovery) do not use raft::server_address to carry IP data
raft: (group0) API refactoring to avoid raft::server_address
raft: rename group0_upgrade.hh to group0_fwd.hh
raft: (group0) move the code around
raft: (discovery) persist a list of discovered peers, not a set
raft: (group0) always start group0 using start_server_for_group0()
Some good news finally. The saved dc/rack info about the ring is now
only loaded once on start. So the whole cache is not needed and the
loading code in storage_service can be greatly simplified
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The l._d._h.'s way to update system keyspace is not like in other code.
Instead of a dedicated helper on the system_keyspace's side it executes
the insertion query directly with the help of qctx.
Now when the l._d._h. has the weak system keyspace reference it can
execute queries on _it_ rather than on the qctx.
Just like in previous patch, it needs to keep the sys._k.s. weak
reference alive until the query's future resolves.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Compaction manager now has the weak reference on the system keyspace
object and can use it to update its stats. It only needs to take care
and keep the shared pointer until the respective future resolves.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
There's a circular dependency between system_keyspace and database. The
former needs the latter because it needs to execula local requests via
query_processor. The latter needs the former via compaction manager and
large data handler, database depends on both and these too need to
insert their entries into system keyspace.
To cut this loop the compaction manager and large data handler both get
a weak reference on the system keysace. Once system keyspace starts is
activcates this reference via the database call. When system keyspace is
shutdown-ed on stop, it deactivates the reference.
Technically the weak reference is implemented by marking the system_k.s.
object as async_sharded_service, and the "reference" in question is the
shared_from_this() pointer. When compaction manager or large data
handler need to update a system keyspace's table, they both hold an
extra reference on the system keyspace until the entry is committed,
thus making sure that sys._k.s. doesn't stop from under their feet. At
the same time, unplugging the reference on shutdown makes sure that no
new entries update will appear and the system_k.s. will eventually be
released.
It's not a C++ classical reference, because system_keyspace starts after
and stops before database.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Many services out there have one (sometimes called .drain()) that's
called early on stop and that's responsible for prearing the service for
stop -- aborting pending/in-flight fibers and alike.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Now all callers have system_keyspace reference at hand. This removes one
more user of the global qctx object
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
This helper needs system_keyspace reference and using "this" as this
looks natural. Also this de-static-ification makes it possible to put
some sense into the invoke_on_all() call from init_system_keyspace()
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
It doesn't match the real system_keyspace_make() definition and is in
fact not needed, as there's another "real" one in database.hh
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
