mirror of
https://github.com/elastic/elasticsearch.git
synced 2025-04-25 07:37:19 -04:00
4023454483
This commit enables concurrent search execution in the DFS phase, which is going to improve resource usage as well as performance of knn queries which benefit from both concurrent rewrite and collection. We will enable concurrent execution for the query phase in a subsequent commit. While this commit does not introduce parallelism for the query phase, it introduces offloading sequential computation to the newly introduced executor. This is true both for situations where a single slice needs to be searched, as well as scenarios where a specific request does not support concurrency (currently only DFS phase does regardless of the request). Sequential collection is not offloaded only if the request includes aggregations that don't support offloading: composite, nested and cardinality as their post collection method must be executed in the same thread as the collection or we'll trip a lucene assertion that verifies that doc_values are pulled and consumed from the same thread. ## Technical details This commit introduces a secondary executor, used exclusively to execute the concurrent bits of search. The search threads are still the ones that coordinate the search (where the caller search will originate from), but the actual work will be offloaded to the newly introduced executor. We are offloading not only parallel execution but also sequential execution, to make the workload more predictable, as it would be surprising to have bits of search executed in either of the two thread pools. Also, that would introduce the possibility to suddenly run a higher amount of heavy operations overall (some in the caller thread and some in the separate threads), which could overload the system as well as make sizing of thread pools more difficult. Note that fetch, together with other actions, is still executed in the search thread pool. This commit does not make the search thread pool merely a coordinating only thread pool, It does so only for what concerns the IndexSearcher#search operation itself, which is though a big portion of the different phases of search API execution. Given that the searcher blocks waiting for all tasks to be completed, we take a simple approach of introducing a thread pool executor that has the same size as the existing search thread pool but relies on an unbounded queue. This simplifies handling of thread pool queue and rejections. In fact, we'd like to guarantee that the secondary thread pool won't reject, and delegate queuing entirely to the search thread pool which is the entry point for every search operation anyway. The principle behind this is that if you got a slot in the search thread pool, you should be able to complete your search, and rather quickly. As part of this commit we are also introducing the ability to cancel tasks that have not started yet, so that if any task throws an exception, other tasks are prevented from starting needless computation. Relates to #80693 Relates to #90700
213 lines
8.3 KiB
Text
213 lines
8.3 KiB
Text
[[modules-threadpool]]
|
|
=== Thread pools
|
|
|
|
A node uses several thread pools to manage memory consumption.
|
|
Queues associated with many of the thread pools enable pending requests
|
|
to be held instead of discarded.
|
|
|
|
There are several thread pools, but the important ones include:
|
|
|
|
`generic`::
|
|
For generic operations (for example, background node discovery).
|
|
Thread pool type is `scaling`.
|
|
|
|
[[search-threadpool]]
|
|
`search`::
|
|
For coordination of count/search operations at the shard level whose computation
|
|
is offloaded to the search_worker thread pool. Used also by fetch and other search
|
|
related operations Thread pool type is `fixed` with a size of `int((`<<node.processors,
|
|
`# of allocated processors`>>`pass:[ * ]3) / 2) + 1`, and queue_size of `1000`.
|
|
|
|
`search_worker`::
|
|
For the heavy workload of count/search operations that may be executed concurrently
|
|
across segments within the same shard when possible. Thread pool type is `fixed`
|
|
with a size of `int((`<<node.processors, `# of allocated processors`>>`pass:[ * ]3) / 2) + 1`, and unbounded queue_size .
|
|
|
|
[[search-throttled]]`search_throttled`::
|
|
For count/search/suggest/get operations on `search_throttled indices`.
|
|
Thread pool type is `fixed` with a size of `1`, and queue_size of `100`.
|
|
|
|
`search_coordination`::
|
|
For lightweight search-related coordination operations. Thread pool type is
|
|
`fixed` with a size of `(`<<node.processors, `# of allocated processors`>>`) / 2`,
|
|
and queue_size of `1000`.
|
|
|
|
`get`::
|
|
For get operations. Thread pool type is
|
|
`fixed` with a size of `int((`<<node.processors,
|
|
`# of allocated processors`>>`pass:[ * ]3) / 2) + 1`, and queue_size of `1000`.
|
|
|
|
`analyze`::
|
|
For analyze requests. Thread pool type is `fixed` with a size of `1`, queue
|
|
size of `16`.
|
|
|
|
`write`::
|
|
For single-document index/delete/update and bulk requests. Thread pool type
|
|
is `fixed` with a size of <<node.processors, `# of allocated processors`>>,
|
|
queue_size of `10000`. The maximum size for this pool is
|
|
`pass:[1 + ]`<<node.processors, `# of allocated processors`>>.
|
|
|
|
`snapshot`::
|
|
For snapshot/restore operations. Thread pool type is `scaling` with a
|
|
keep-alive of `5m`. On nodes with at least 750MB of heap the maximum size
|
|
of this pool is `10` by default. On nodes with less than 750MB of heap the
|
|
maximum size of this pool is `min(5, (`<<node.processors,
|
|
`# of allocated processors`>>`) / 2)` by default.
|
|
|
|
`snapshot_meta`::
|
|
For snapshot repository metadata read operations. Thread pool type is `scaling` with a
|
|
keep-alive of `5m` and a max of `min(50, (`<<node.processors,
|
|
`# of allocated processors`>>`* 3))`.
|
|
|
|
`warmer`::
|
|
For segment warm-up operations. Thread pool type is `scaling` with a
|
|
keep-alive of `5m` and a max of `min(5, (`<<node.processors,
|
|
`# of allocated processors`>>`) / 2)`.
|
|
|
|
`refresh`::
|
|
For refresh operations. Thread pool type is `scaling` with a
|
|
keep-alive of `5m` and a max of `min(10, (`<<node.processors,
|
|
`# of allocated processors`>>`) / 2)`.
|
|
|
|
`fetch_shard_started`::
|
|
For listing shard states.
|
|
Thread pool type is `scaling` with keep-alive of `5m` and a default maximum
|
|
size of `pass:[2 * ]`<<node.processors, `# of allocated processors`>>.
|
|
|
|
`fetch_shard_store`::
|
|
For listing shard stores.
|
|
Thread pool type is `scaling` with keep-alive of `5m` and a default maximum
|
|
size of `pass:[2 * ]`<<node.processors, `# of allocated processors`>>.
|
|
|
|
`flush`::
|
|
For <<indices-flush,flush>> and <<index-modules-translog, translog>> `fsync`
|
|
operations. Thread pool type is `scaling` with a keep-alive of `5m` and a
|
|
default maximum size of `min(5, (`<<node.processors,
|
|
`# of allocated processors`>>`) / 2)`.
|
|
|
|
`force_merge`::
|
|
For <<indices-forcemerge,force merge>> operations.
|
|
Thread pool type is `fixed` with a size of `max(1, (`<<node.processors,
|
|
`# of allocated processors`>>`) / 8)` and an unbounded queue size.
|
|
|
|
`management`::
|
|
For cluster management.
|
|
Thread pool type is `scaling` with a keep-alive of `5m` and a default
|
|
maximum size of `5`.
|
|
|
|
`system_read`::
|
|
For read operations on system indices.
|
|
Thread pool type is `fixed` with a default maximum size of
|
|
`min(5, (`<<node.processors, `# of allocated processors`>>`) / 2)`.
|
|
|
|
`system_write`::
|
|
For write operations on system indices.
|
|
Thread pool type is `fixed` with a default maximum size of
|
|
`min(5, (`<<node.processors, `# of allocated processors`>>`) / 2)`.
|
|
|
|
`system_critical_read`::
|
|
For critical read operations on system indices.
|
|
Thread pool type is `fixed` with a default maximum size of
|
|
`min(5, (`<<node.processors, `# of allocated processors`>>`) / 2)`.
|
|
|
|
`system_critical_write`::
|
|
For critical write operations on system indices.
|
|
Thread pool type is `fixed` with a default maximum size of
|
|
`min(5, (`<<node.processors, `# of allocated processors`>>`) / 2)`.
|
|
|
|
`watcher`::
|
|
For <<xpack-alerting,watch executions>>.
|
|
Thread pool type is `fixed` with a default maximum size of
|
|
`min(5 * (`<<node.processors, `# of allocated processors`>>`), 50)`
|
|
and queue_size of `1000`.
|
|
|
|
Thread pool settings are <<static-cluster-setting,static>> and can be changed by
|
|
editing `elasticsearch.yml`. Changing a specific thread pool can be done by
|
|
setting its type-specific parameters; for example, changing the number of
|
|
threads in the `write` thread pool:
|
|
|
|
[source,yaml]
|
|
--------------------------------------------------
|
|
thread_pool:
|
|
write:
|
|
size: 30
|
|
--------------------------------------------------
|
|
|
|
[[thread-pool-types]]
|
|
==== Thread pool types
|
|
|
|
The following are the types of thread pools and their respective parameters:
|
|
|
|
[[fixed-thread-pool]]
|
|
===== `fixed`
|
|
|
|
The `fixed` thread pool holds a fixed size of threads to handle the
|
|
requests with a queue (optionally bounded) for pending requests that
|
|
have no threads to service them.
|
|
|
|
The `size` parameter controls the number of threads.
|
|
|
|
The `queue_size` allows to control the size of the queue of pending
|
|
requests that have no threads to execute them. By default, it is set to
|
|
`-1` which means its unbounded. When a request comes in and the queue is
|
|
full, it will abort the request.
|
|
|
|
[source,yaml]
|
|
--------------------------------------------------
|
|
thread_pool:
|
|
write:
|
|
size: 30
|
|
queue_size: 1000
|
|
--------------------------------------------------
|
|
|
|
[[scaling-thread-pool]]
|
|
===== `scaling`
|
|
|
|
The `scaling` thread pool holds a dynamic number of threads. This
|
|
number is proportional to the workload and varies between the value of
|
|
the `core` and `max` parameters.
|
|
|
|
The `keep_alive` parameter determines how long a thread should be kept
|
|
around in the thread pool without it doing any work.
|
|
|
|
[source,yaml]
|
|
--------------------------------------------------
|
|
thread_pool:
|
|
warmer:
|
|
core: 1
|
|
max: 8
|
|
keep_alive: 2m
|
|
--------------------------------------------------
|
|
|
|
[[node.processors]]
|
|
==== Allocated processors setting
|
|
|
|
The number of processors is automatically detected, and the thread pool settings
|
|
are automatically set based on it. In some cases it can be useful to override
|
|
the number of detected processors. This can be done by explicitly setting the
|
|
`node.processors` setting. This setting is bounded by the number of available
|
|
processors and accepts floating point numbers, which can be useful in environments
|
|
where the {es} nodes are configured to run with CPU limits, such as cpu
|
|
shares or quota under `Cgroups`.
|
|
|
|
[source,yaml]
|
|
--------------------------------------------------
|
|
node.processors: 2
|
|
--------------------------------------------------
|
|
|
|
There are a few use-cases for explicitly overriding the `node.processors`
|
|
setting:
|
|
|
|
. If you are running multiple instances of {es} on the same host but want
|
|
{es} to size its thread pools as if it only has a fraction of the CPU, you
|
|
should override the `node.processors` setting to the desired fraction, for
|
|
example, if you're running two instances of {es} on a 16-core machine, set
|
|
`node.processors` to 8. Note that this is an expert-level use case and there's
|
|
a lot more involved than just setting the `node.processors` setting as there are
|
|
other considerations like changing the number of garbage collector threads,
|
|
pinning processes to cores, and so on.
|
|
. Sometimes the number of processors is wrongly detected and in such cases
|
|
explicitly setting the `node.processors` setting will workaround such issues.
|
|
|
|
In order to check the number of processors detected, use the nodes info
|
|
API with the `os` flag.
|