Automatic optimisation for ingest chunking based on processed messages

This commit is contained in:
Mark Veidemanis 2025-01-23 11:29:48 +00:00
parent 1aeadaf3b7
commit 1cc2ef629e
Signed by: m
GPG Key ID: 5ACFCEED46C0904F
3 changed files with 85 additions and 61 deletions

View File

@ -2,17 +2,14 @@ version: "2.2"
services:
app:
image: pathogen/monolith:latest
image: xf/monolith:latest
container_name: monolith
build: .
volumes:
- ${PORTAINER_GIT_DIR}:/code
# env_file:
# - stack.env
networks:
- default
- xf
- db
- type: bind
source: /code/run
target: /var/run
environment:
PORTAINER_GIT_DIR: "${PORTAINER_GIT_DIR}"
MODULES_ENABLED: "${MODULES_ENABLED}"
@ -49,41 +46,22 @@ services:
MONOLITH_PROCESS_PERFSTATS: "${MONOLITH_PROCESS_PERFSTATS}"
MONOLITH_CH4_BOARDS: "${MONOLITH_CH4_BOARDS}"
REDIS_PASSWORD: "${REDIS_PASSWORD}"
MONOLITH_INGEST_INCREASE_BELOW: "${MONOLITH_INGEST_INCREASE_BELOW}"
MONOLITH_INGEST_INCREASE_BY: "${MONOLITH_INGEST_INCREASE_BY}"
MONOLITH_INGEST_DECREASE_ABOVE: "${MONOLITH_INGEST_DECREASE_ABOVE}"
MONOLITH_INGEST_DECREASE_BY: "${MONOLITH_INGEST_DECREASE_BY}"
MONOLITH_INGEST_MAX: "${MONOLITH_INGEST_MAX}"
MONOLITH_INGEST_MIN: "${MONOLITH_INGEST_MIN}"
deploy:
resources:
limits:
cpus: '0.5'
memory: 1.0G
network_mode: host
db:
#image: pathogen/manticore:kibana
image: manticoresearch/manticore:dev
container_name: monolith_db
#build:
# context: ./docker/manticore
# args:
# DEV: 1
restart: always
ports:
- 9308
- 9312
- 9306
ulimits:
nproc: 65535
nofile:
soft: 65535
hard: 65535
memlock:
soft: -1
hard: -1
environment:
# - MCL=1
- EXTRA=1
networks:
- default
- xf
- db
volumes:
- ./docker/data:/var/lib/manticore
# - ./docker/manticore.conf:/etc/manticoresearch/manticore.conf
# threshold:
# image: pathogen/threshold:latest
# image: xf/threshold:latest
# container_name: threshold
# build: legacy/docker
# volumes:
@ -148,16 +126,17 @@ services:
- "1289:1289"
environment:
- SSDB_PORT=1289
networks:
- default
- db
# tmp:
# image: busybox
# container_name: tmp_monolith
# command: chmod -R 777 /var/run/socks
# volumes:
# - /var/run/socks
volumes:
- monolith_ssdb_data:/var/lib/ssdb
# networks:
# - default
# - db
deploy:
resources:
limits:
cpus: '0.5'
memory: 1.0G
network_mode: host
redis:
image: redis
@ -171,6 +150,9 @@ services:
volumes:
- ${PORTAINER_GIT_DIR}/docker/redis.conf:/etc/redis.conf
- monolith_redis_data:/data
- type: bind
source: /code/run
target: /var/run
# volumes_from:
# - tmp
healthcheck:
@ -178,18 +160,25 @@ services:
interval: 2s
timeout: 2s
retries: 15
networks:
- default
- xf
- db
# networks:
# - default
# - xf
# - db
deploy:
resources:
limits:
cpus: '0.5'
memory: 1.0G
network_mode: host
networks:
default:
driver: bridge
xf:
external: true
db:
external: true
# networks:
# default:
# driver: bridge
# xf:
# external: true
# db:
# external: true
volumes:
monolith_redis_data:
monolith_ssdb_data:

View File

@ -22,3 +22,4 @@ orjson
uvloop
elasticsearch[async]
msgpack
# flpc

View File

@ -11,8 +11,15 @@ KEYNAME = "queue"
CHUNK_SIZE = int(getenv("MONOLITH_INGEST_CHUNK_SIZE", "900"))
ITER_DELAY = float(getenv("MONOLITH_INGEST_ITER_DELAY", "0.5"))
INGEST_INCREASE_BELOW = int(getenv("MONOLITH_INGEST_INCREASE_BELOW", "2500"))
INGEST_DECREASE_ABOVE = int(getenv("MONOLITH_INGEST_DECREASE_ABOVE", "10000"))
INGEST_INCREASE_BY = int(getenv("MONOLITH_INGEST_INCREASE_BY", "100"))
INGEST_DECREASE_BY = int(getenv("MONOLITH_INGEST_DECREASE_BY", "100"))
log = util.get_logger("ingest")
INGEST_MAX = int(getenv("MONOLITH_INGEST_MAX", "1000000"))
INGEST_MIN = int(getenv("MONOLITH_INGEST_MIN", "100"))
class Ingest(object):
def __init__(self):
@ -34,9 +41,36 @@ class Ingest(object):
await asyncio.sleep(ITER_DELAY)
async def get_chunk(self):
global CHUNK_SIZE
length = await db.ar.llen(KEYNAME)
if length > CHUNK_SIZE:
length = CHUNK_SIZE
if not length:
return
await process.spawn_processing_threads(self.current_chunk, length)
ingested = await process.spawn_processing_threads(self.current_chunk, length)
if ingested < INGEST_INCREASE_BELOW:
if CHUNK_SIZE + INGEST_INCREASE_BY < INGEST_MAX:
self.log.info(
(
f"Increasing chunk size to "
f"{CHUNK_SIZE + INGEST_INCREASE_BY} "
f"due to low ingestion ({ingested})"
)
)
CHUNK_SIZE += INGEST_INCREASE_BY
else:
log.info(f"Chunk size ({CHUNK_SIZE}) at maximum, not increasing above: {INGEST_MAX}")
elif ingested > INGEST_DECREASE_ABOVE:
if CHUNK_SIZE - INGEST_DECREASE_BY > INGEST_MIN:
self.log.info(
(
f"Decreasing chunk size to "
f"{CHUNK_SIZE - INGEST_DECREASE_BY}"
f"due to high ingestion ({ingested})"
)
)
CHUNK_SIZE -= INGEST_DECREASE_BY
else:
log.info(f"Chunk size ({CHUNK_SIZE}) at minimum, not decreasing below: {INGEST_MIN}")