From 523f71576edce05501ab146ceddd78311f24279a Mon Sep 17 00:00:00 2001
From: Georgy Moshkin <gmoshkin@picodata.io>
Date: Wed, 20 Mar 2024 13:52:26 +0300
Subject: [PATCH] feat: config.yaml: add support for some memtx, vinyl & iproto
box.cfg parameters
---
src/config.rs | 267 ++++++++++++++++++++++++++++++++++-
src/tarantool.rs | 13 ++
test/int/test_config_file.py | 67 ++++++++-
3 files changed, 338 insertions(+), 9 deletions(-)
diff --git a/src/config.rs b/src/config.rs
index 8939454395..8bbaec8fc6 100644
--- a/src/config.rs
+++ b/src/config.rs
@@ -247,7 +247,7 @@ Using configuration file '{args_path}'.");
}
if let Some(memtx_memory) = args.memtx_memory {
- self.instance.memtx_memory = Some(memtx_memory);
+ self.instance.memtx.memory = Some(memtx_memory);
}
// --config-parameter has higher priority than other command line
@@ -623,12 +623,22 @@ pub struct InstanceConfig {
/// deleting.
pub shredding: Option<bool>,
- pub memtx_memory: Option<u64>,
-
#[serde(default)]
#[introspection(nested)]
pub log: LogSection,
+ #[serde(default)]
+ #[introspection(nested)]
+ pub memtx: MemtxSection,
+
+ #[serde(default)]
+ #[introspection(nested)]
+ pub vinyl: VinylSection,
+
+ #[serde(default)]
+ #[introspection(nested)]
+ pub iproto: IprotoSection,
+
/// Special catch-all field which will be filled by serde with all unknown
/// fields from the yaml file.
#[serde(flatten)]
@@ -721,10 +731,255 @@ impl InstanceConfig {
#[inline]
pub fn memtx_memory(&self) -> u64 {
- self.memtx_memory.unwrap_or(64 * 1024 * 1024)
+ self.memtx.memory.unwrap_or(64 * 1024 * 1024)
+ }
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// MemtxSection
+////////////////////////////////////////////////////////////////////////////////
+
+#[derive(
+ PartialEq,
+ Default,
+ Debug,
+ Clone,
+ serde::Deserialize,
+ serde::Serialize,
+ tlua::Push,
+ tlua::PushInto,
+ Introspection,
+)]
+#[serde(deny_unknown_fields)]
+pub struct MemtxSection {
+ /// Enable [transactional manager](https://www.tarantool.io/en/doc/latest/concepts/atomic/txn_mode_mvcc/#txn-mode-transaction-manager)
+ /// if set to true.
+ // pub use_mvcc_engine: Option<bool>,
+
+ /// Specify the allocator that manages memory for memtx tuples. Possible values:
+ /// - `system` – the memory is allocated as needed, checking that the quota
+ /// is not exceeded. The allocator is based on the `malloc` function.
+ /// - `small` – a slab allocator. The allocator repeatedly uses a memory
+ /// block to allocate objects of the same type. Note that this allocator is
+ /// prone to unresolvable fragmentation on specific workloads, so you can
+ /// switch to system in such cases.
+ ///
+ /// Corresponds to `box.cfg.memtx_allocator`.
+ // pub allocator: Option<MemtxAllocator>,
+
+ /// The multiplier for computing the sizes of memory chunks that tuples are
+ /// stored in. A lower value may result in less wasted memory depending on
+ /// the total amount of memory available and the distribution of item sizes.
+ ///
+ /// Corresponds to `box.cfg.slab_alloc_factor`.
+ pub allocation_factor: Option<f64>,
+
+ /// Specify the granularity (in bytes) of memory allocation in the small
+ /// allocator. The memtx.slab_alloc_granularity value should meet the
+ /// following conditions:
+ /// - The value is a power of two.
+ /// - The value is greater than or equal to 4.
+ ///
+ /// Below are few recommendations on how to adjust the
+ /// memtx.slab_alloc_granularity option:
+ /// - If the tuples in space are small and have about the same size, set the
+ /// option to 4 bytes to save memory.
+ /// - If the tuples are different-sized, increase the option value to
+ /// allocate tuples from the same mempool (memory pool).
+ ///
+ /// Corresponds to `box.cfg.slab_alloc_granularity`.
+ // pub allocation_granularity: Option<u64>,
+ // Not supported yet.
+ // Size of the smallest allocation unit. It can be decreased if most of the
+ // tuples are very small.
+ // pub min_tuple_size: Option<u64>,
+
+ // Not supported yet.
+ // Size of the largest allocation unit, for the memtx storage engine. It can
+ // be increased if it is necessary to store large tuples.
+ // pub max_tuple_size: Option<u64>,
+
+ /// How much memory is allocated to store tuples. When the limit is
+ /// reached, INSERT or UPDATE requests begin failing with error
+ /// ER_MEMORY_ISSUE. The server does not go beyond the memtx_memory limit to
+ /// allocate tuples, but there is additional memory used to store indexes
+ /// and connection information.
+ ///
+ /// Minimum is 32MB (32 * 1024 * 1024).
+ ///
+ /// Corresponds to `box.cfg.memtx_memory`.
+ pub memory: Option<u64>,
+
+ /// The maximum number of snapshots that are stored in the memtx_dir
+ /// directory. If the number of snapshots after creating a new one exceeds
+ /// this value, the Tarantool garbage collector deletes old snapshots. If
+ /// the option is set to zero, the garbage collector does not delete old
+ /// snapshots.
+ ///
+ /// Corresponds to `box.cfg.checkpoint_count`.
+ pub checkpoint_count: Option<u64>,
+
+ /// The interval in seconds between actions by the checkpoint daemon. If the
+ /// option is set to a value greater than zero, and there is activity that
+ /// causes change to a database, then the checkpoint daemon calls
+ /// box.snapshot() every checkpoint_interval seconds, creating a new
+ /// snapshot file each time. If the option is set to zero, the checkpoint
+ /// daemon is disabled.
+ ///
+ /// Corresponds to `box.cfg.checkpoint_interval`.
+ pub checkpoint_interval: Option<f64>,
+}
+
+tarantool::define_str_enum! {
+ #[derive(Default)]
+ pub enum MemtxAllocator {
+ #[default]
+ Small = "small",
+ System = "system",
}
}
+////////////////////////////////////////////////////////////////////////////////
+// VinylSection
+////////////////////////////////////////////////////////////////////////////////
+
+#[derive(
+ PartialEq,
+ Default,
+ Debug,
+ Clone,
+ serde::Deserialize,
+ serde::Serialize,
+ tlua::Push,
+ tlua::PushInto,
+ Introspection,
+)]
+#[serde(deny_unknown_fields)]
+pub struct VinylSection {
+ /// The maximum number of in-memory bytes that vinyl uses.
+ ///
+ /// Corresponds to `box.cfg.vinyl_memory`
+ pub memory: Option<u64>,
+
+ /// The cache size for the vinyl storage engine.
+ ///
+ /// Corresponds to `box.cfg.vinyl_cache`
+ pub cache: Option<u64>,
+
+ /// The maximum number of read threads that vinyl can use for some
+ /// concurrent operations, such as I/O and compression.
+ ///
+ /// Corresponds to `box.cfg.vinyl_read_threads`
+ pub read_threads: Option<u64>,
+
+ /// The maximum number of write threads that vinyl can use for some
+ /// concurrent operations, such as I/O and compression.
+ ///
+ /// Corresponds to `box.cfg.vinyl_write_threads`
+ pub write_threads: Option<u64>,
+
+ // pub max_tuple_size: Option<u64>, <- не надо Ñто разрешать пока,
+ /// Enables the deferred DELETE optimization for vinyl spaces by default.
+ ///
+ /// This can also be controlled on a per-table basis in the options for
+ /// `space_object:create_index()` (we don't support this yet in picodata).
+ ///
+ /// Corresponds to `box.cfg.vinyl_defer_deletes`
+ pub default_defer_deletes: Option<bool>,
+
+ /// Page size. Page is a read/write unit for vinyl disk operations.
+ ///
+ /// This can also be controlled on a per-table basis in the options for
+ /// `space_object:create_index()` (we don't support this yet in picodata).
+ ///
+ /// Corresponds to `box.cfg.vinyl_page_size`
+ pub default_page_size: Option<u64>,
+
+ /// The maximal number of runs per level in vinyl LSM tree.
+ /// If this number is exceeded, a new level is created.
+ ///
+ /// This can also be controlled on a per-table basis in the options for
+ /// `space_object:create_index()` (we don't support this yet in picodata).
+ ///
+ /// Corresponds to `box.cfg.vinyl_run_count_per_level`
+ pub default_run_count_per_level: Option<u64>,
+
+ /// Ratio between the sizes of different levels in the LSM tree.
+ ///
+ /// This can also be controlled on a per-table basis in the options for
+ /// `space_object:create_index()` (we don't support this yet in picodata).
+ ///
+ /// Corresponds to `box.cfg.vinyl_run_size_ratio`
+ pub default_run_size_ratio: Option<f64>,
+
+ /// Bloom filter false positive rate – the suitable probability of the bloom
+ /// filter to give a wrong result.
+ ///
+ /// This can also be controlled on a per-table basis in the options for
+ /// `space_object:create_index()` (we don't support this yet in picodata).
+ ///
+ /// Corresponds to `box.cfg.vinyl_bloom_fpr`
+ pub default_bloom_fpr: Option<f64>,
+ // pub vinyl_timeout: Option<f64>, // do we need this also?
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// IprotoSection
+////////////////////////////////////////////////////////////////////////////////
+
+#[derive(
+ PartialEq,
+ Default,
+ Debug,
+ Clone,
+ serde::Deserialize,
+ serde::Serialize,
+ tlua::Push,
+ tlua::PushInto,
+ Introspection,
+)]
+#[serde(deny_unknown_fields)]
+pub struct IprotoSection {
+ /// To handle messages, Tarantool allocates fibers. To prevent fiber
+ /// overhead from affecting the whole system, Tarantool restricts how many
+ /// messages the fibers handle, so that some pending requests are blocked.
+ ///
+ /// On powerful systems, increase net_msg_max and the scheduler will
+ /// immediately start processing pending requests.
+ ///
+ /// On weaker systems, decrease net_msg_max and the overhead may decrease
+ /// although this may take some time because the scheduler must wait until
+ /// already-running requests finish.
+ ///
+ /// When net_msg_max is reached, Tarantool suspends processing of incoming
+ /// packages until it has processed earlier messages. This is not a direct
+ /// restriction of the number of fibers that handle network messages, rather
+ /// it is a system-wide restriction of channel bandwidth. This in turn
+ /// causes restriction of the number of incoming network messages that the
+ /// transaction processor thread handles, and therefore indirectly affects
+ /// the fibers that handle network messages. (The number of fibers is
+ /// smaller than the number of messages because messages can be released as
+ /// soon as they are delivered, while incoming requests might not be
+ /// processed until some time after delivery.)
+ ///
+ /// On typical systems, the default value (768) is correct.
+ ///
+ /// Corresponds to `box.cfg.net_msg_max`
+ pub max_concurrent_messages: Option<u64>,
+ // /// The size of the read-ahead buffer associated with a client connection.
+ // /// The larger the buffer, the more memory an active connection consumes and
+ // /// the more requests can be read from the operating system buffer in a
+ // /// single system call. The rule of thumb is to make sure the buffer can
+ // /// contain at least a few dozen requests. Therefore, if a typical tuple in
+ // /// a request is large, e.g. a few kilobytes or even megabytes, the
+ // /// read-ahead buffer size should be increased. If batched request
+ // /// processing is not used, it’s prudent to leave this setting at its
+ // /// default.
+ // ///
+ // /// Corresponds to `box.cfg.readahead`
+ // pub readahead_buffer_size: Option<u64>,
+}
+
////////////////////////////////////////////////////////////////////////////////
// LogSection
////////////////////////////////////////////////////////////////////////////////
@@ -1343,13 +1598,13 @@ instance:
let mut config = PicodataConfig::read_yaml_contents(&yaml).unwrap();
let args = args::Run::try_parse_from(["run",
"-c", " instance.log .level =debug ",
- "--config-parameter", "instance. memtx_memory= 0xdeadbeef",
+ "--config-parameter", "instance. memtx . memory= 0xdeadbeef",
]).unwrap();
config.set_from_args(args).unwrap();
assert_eq!(config.instance.tier.unwrap(), "ABC");
assert_eq!(config.cluster.cluster_id.unwrap(), "DEF");
assert_eq!(config.instance.log.level.unwrap(), args::LogLevel::Debug);
- assert_eq!(config.instance.memtx_memory.unwrap(), 0xdead_beef);
+ assert_eq!(config.instance.memtx.memory.unwrap(), 0xdead_beef);
assert_eq!(config.instance.audit.unwrap(), "audit.txt");
assert_eq!(config.instance.data_dir.unwrap(), ".");
diff --git a/src/tarantool.rs b/src/tarantool.rs
index 4642fe4194..347b5c413f 100644
--- a/src/tarantool.rs
+++ b/src/tarantool.rs
@@ -223,6 +223,19 @@ impl Cfg {
const MAPPING: &[(&str, &str)] = &[
// Other instance.log.* parameters are set explicitly above
("log_format", "instance.log.format"),
+ ("slab_alloc_factor", "instance.memtx.allocation_factor"),
+ ("checkpoint_count", "instance.memtx.checkpoint_count"),
+ ("checkpoint_interval", "instance.memtx.checkpoint_interval"),
+ ("vinyl_memory", "instance.vinyl.memory"),
+ ("vinyl_cache", "instance.vinyl.cache"),
+ ("vinyl_read_threads", "instance.vinyl.read_threads"),
+ ("vinyl_write_threads", "instance.vinyl.write_threads"),
+ ("vinyl_defer_deletes", "instance.vinyl.default_defer_deletes"),
+ ("vinyl_page_size", "instance.vinyl.default_page_size"),
+ ("vinyl_run_count_per_level", "instance.vinyl.default_run_count_per_level"),
+ ("vinyl_run_size_ratio", "instance.vinyl.default_run_size_ratio"),
+ ("vinyl_bloom_fpr", "instance.vinyl.default_bloom_fpr"),
+ ("net_msg_max", "instance.iproto.max_concurrent_messages"),
];
for (box_field, picodata_field) in MAPPING {
let value = config
diff --git a/test/int/test_config_file.py b/test/int/test_config_file.py
index 2bf975f670..0961c4059f 100644
--- a/test/int/test_config_file.py
+++ b/test/int/test_config_file.py
@@ -12,7 +12,9 @@ instance:
cluster_id: test
instance_id: from-config
replicaset_id: with-love
- memtx_memory: 42069
+
+ memtx:
+ memory: 42069
"""
)
instance = cluster.add_instance(instance_id=False, wait_online=False)
@@ -90,6 +92,9 @@ instance:
listen=dict(host=host, port=str(port)),
log=dict(level="verbose"),
peers=[dict(host=host, port=str(port))],
+ iproto=[],
+ memtx=[],
+ vinyl=[],
unknown_parameters=[],
),
unknown_sections=[],
@@ -110,7 +115,8 @@ cluster:
tiers:
default:
instance:
- memtx_memory: 0xdeadbeef
+ memtx:
+ memory: 0xdeadbeef
"""
)
instance.start(cwd=work_dir)
@@ -129,7 +135,8 @@ cluster:
tiers:
default:
instance:
- memtx_memory: 0xcafebabe
+ memtx:
+ memory: 0xcafebabe
"""
)
instance.env["PICODATA_CONFIG_FILE"] = config_path
@@ -270,6 +277,23 @@ cluster:
assert box_cfg["log_level"] == 6 # means verbose -- set by our testing harness
assert box_cfg["log_format"] == "plain"
+ assert box_cfg["memtx_memory"] == 64 * 1024 * 1024
+ assert box_cfg["slab_alloc_factor"] == 1.05
+ assert box_cfg["checkpoint_count"] == 2
+ assert box_cfg["checkpoint_interval"] == 3600
+
+ assert box_cfg["vinyl_memory"] == 128 * 1024 * 1024
+ assert box_cfg["vinyl_cache"] == 128 * 1024 * 1024
+ assert box_cfg["vinyl_read_threads"] == 1
+ assert box_cfg["vinyl_write_threads"] == 4
+ assert box_cfg["vinyl_defer_deletes"] == False # noqa: E712
+ assert box_cfg["vinyl_page_size"] == 8 * 1024
+ assert box_cfg["vinyl_run_count_per_level"] == 2
+ assert box_cfg["vinyl_run_size_ratio"] == 3.5
+ assert box_cfg["vinyl_bloom_fpr"] == 0.05
+
+ assert box_cfg["net_msg_max"] == 0x300
+
#
# Check explicitly set values
#
@@ -286,6 +310,26 @@ instance:
destination: file:/proc/self/fd/2 # this is how you say `stderr` explicitly
level: debug
format: json
+
+ memtx:
+ memory: 0x7777777
+ allocation_factor: 1.7
+ checkpoint_count: 8
+ checkpoint_interval: 1800
+
+ vinyl:
+ memory: 0x8888888
+ cache: 0x4444444
+ read_threads: 2
+ write_threads: 3
+ default_defer_deletes: true
+ default_page_size: 0x8000
+ default_run_count_per_level: 3
+ default_run_size_ratio: 7
+ default_bloom_fpr: 0.777
+
+ iproto:
+ max_concurrent_messages: 0x600
"""
)
@@ -301,3 +345,20 @@ instance:
assert box_cfg["log"] == "file:/proc/self/fd/2"
assert box_cfg["log_level"] == 7 # means debug
assert box_cfg["log_format"] == "json"
+
+ assert box_cfg["memtx_memory"] == 0x777_7777
+ assert box_cfg["slab_alloc_factor"] == 1.7
+ assert box_cfg["checkpoint_count"] == 8
+ assert box_cfg["checkpoint_interval"] == 1800
+
+ assert box_cfg["vinyl_memory"] == 0x8888888
+ assert box_cfg["vinyl_cache"] == 0x4444444
+ assert box_cfg["vinyl_read_threads"] == 2
+ assert box_cfg["vinyl_write_threads"] == 3
+ assert box_cfg["vinyl_defer_deletes"] == True # noqa: E712
+ assert box_cfg["vinyl_page_size"] == 0x8000
+ assert box_cfg["vinyl_run_count_per_level"] == 3
+ assert box_cfg["vinyl_run_size_ratio"] == 7
+ assert box_cfg["vinyl_bloom_fpr"] == 0.777
+
+ assert box_cfg["net_msg_max"] == 0x600
--
GitLab