milvus/cpp/unittest/server/appendix/server_config.yaml
starlord 5c36b1a052 refine cache code
Former-commit-id: 2a982ed5663a5e34c46d8a3be2bb5c875ee6b63a
2019-09-16 15:46:58 +08:00

46 lines
2.6 KiB
YAML

server_config:
address: 0.0.0.0 # milvus server ip address (IPv4)
port: 19530 # the port milvus listen to, default: 19530, range: 1025 ~ 65534
mode: single # milvus deployment type: single, cluster, read_only
time_zone: UTC+8 # Use the UTC-x or UTC+x to specify a time zone. eg. UTC+8 for China Standard Time
db_config:
db_path: /tmp/milvus # milvus data storage path
db_slave_path: # secondry data storage path, split by semicolon
# URI format: dialect://username:password@host:port/database
# All parts except dialect are optional, but you MUST include the delimiters
# Currently dialect supports mysql or sqlite
db_backend_url: sqlite://:@:/
archive_disk_threshold: 0 # triger archive action if storage size exceed this value, 0 means no limit, unit: GB
archive_days_threshold: 0 # files older than x days will be archived, 0 means no limit, unit: day
insert_buffer_size: 4 # maximum insert buffer size allowed, default: 4, unit: GB, should be at least 1 GB.
# the sum of insert_buffer_size and cpu_cache_capacity should be less than total memory, unit: GB
build_index_gpu: 0 # which gpu is used to build index, default: 0, range: 0 ~ gpu number - 1
metric_config:
is_startup: off # if monitoring start: on, off
collector: prometheus # metrics collector: prometheus
prometheus_config: # following are prometheus configure
port: 8080 # the port prometheus use to fetch metrics
push_gateway_ip_address: 127.0.0.1 # push method configure: push gateway ip address
push_gateway_port: 9091 # push method configure: push gateway port
cache_config:
cpu_cache_capacity: 16 # how many memory are used as cache, unit: GB, range: 0 ~ less than total memory
cpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
insert_cache_immediately: false # insert data will be load into cache immediately for hot query
gpu_cache_capacity: 5 # how many memory are used as cache in gpu, unit: GB, RANGE: 0 ~ less than total memory
gpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
engine_config:
use_blas_threshold: 20
resource_config:
mode: simple
resources:
# - cpu
- gpu0