Related configuration of queryNode, used to run hybrid search between vector and scalar data.
| Description |
Default Value |
| The interval that query node publishes the node statistics information, including segment status, cpu usage, memory usage, health status, etc. Unit: ms. |
1000 |
| Description |
Default Value |
| The number of threads in knowhere's thread pool. If disk is enabled, the pool size will multiply with knowhereThreadPoolNumRatio([1, 32]). |
4 |
| Description |
Default Value |
| Row count by which Segcore divides a segment into chunks. |
128 |
| Description |
Default Value |
|
Whether to create a temporary index for growing segments and sealed segments not yet indexed, improving search performance.
Milvus will eventually seals and indexes all segments, but enabling this optimizes search performance for immediate queries following data insertion.
This defaults to true, indicating that Milvus creates temporary index for growing segments and the sealed segments that are not indexed upon searches. |
true |
| Description |
Default Value |
| temp index nlist, recommend to set sqrt(chunkRows), must smaller than chunkRows/8 |
128 |
| Description |
Default Value |
| nprobe to search small index, based on your accuracy requirement, must smaller than nlist |
16 |
| Description |
Default Value |
| extra memory needed by building interim index |
1.15 |
| Description |
Default Value |
| the ratio of building interim index parallel matched with cpu num |
0.5 |
| Description |
Default Value |
| Enable multiple chunked search |
true |
| Description |
Default Value |
| Enable knowhere strong consistency score computation logic |
false |
| Description |
Default Value |
| The multiply factor of calculating the memory usage while loading segments |
1 |
| Description |
Default Value |
| enable querynode load disk index, and search on disk index |
false |
| Description |
Default Value |
| 2 GB, 2 * 1024 *1024 *1024 |
2147483648 |
| Description |
Default Value |
| The read ahead policy of chunk cache, options: `normal, random, sequential, willneed, dontneed` |
willneed |
| Description |
Default Value |
|
options: async, sync, disable.
Specifies the necessity for warming up the chunk cache.
1. If set to "sync" or "async" the original vector data will be synchronously/asynchronously loaded into the
chunk cache during the load process. This approach has the potential to substantially reduce query/search latency
for a specific duration post-load, albeit accompanied by a concurrent increase in disk usage;
2. If set to "disable" original vector data will only be loaded into the chunk cache during search/query. |
disable |
| Description |
Default Value |
| Enable mmap for loading vector data |
false |
| Description |
Default Value |
| Enable mmap for loading vector index |
false |
| Description |
Default Value |
| Enable mmap for loading scalar data |
false |
| Description |
Default Value |
| Enable mmap for loading scalar index |
false |
| Description |
Default Value |
| Enable mmap for chunk cache (raw vector retrieving). |
true |
| Description |
Default Value |
|
Enable memory mapping (mmap) to optimize the handling of growing raw data.
By activating this feature, the memory overhead associated with newly added or modified data will be significantly minimized.
However, this optimization may come at the cost of a slight decrease in query latency for the affected data segments. |
false |
| Description |
Default Value |
| Enable lazyload for loading data |
false |
| Description |
Default Value |
| max wait timeout duration in milliseconds before start to do lazyload search and retrieve |
30000 |
| Description |
Default Value |
| max timeout in milliseconds for waiting request resource for lazy load, 5s by default |
5000 |
| Description |
Default Value |
| retry interval in milliseconds for waiting request resource for lazy load, 2s by default |
2000 |
| Description |
Default Value |
| max retry times for lazy load, 1 by default |
1 |
| Description |
Default Value |
| max evict count for lazy load, 1 by default |
1 |
| Description |
Default Value |
| enable index offset cache for some scalar indexes, now is just for bitmap index, enable this param can improve performance for retrieving raw data from index |
false |
| Description |
Default Value |
|
maxReadConcurrentRatio is the concurrency ratio of read task (search task and query task).
Max read concurrency would be the value of hardware.GetCPUNum * maxReadConcurrentRatio.
It defaults to 2.0, which means max read concurrency would be the value of hardware.GetCPUNum * 2.
Max read concurrency must greater than or equal to 1, and less than or equal to hardware.GetCPUNum * 100.
(0, 100] |
1 |
| Description |
Default Value |
| ratio used to estimate read task cpu usage. |
10 |
| Description |
Default Value |
|
fifo: A FIFO queue support the schedule.
user-task-polling:
The user's tasks will be polled one by one and scheduled.
Scheduling is fair on task granularity.
The policy is based on the username for authentication.
And an empty username is considered the same user.
When there are no multi-users, the policy decay into FIFO" |
fifo |
| Description |
Default Value |
| Control how long (many seconds) that queue retains since queue is empty |
60 |
| Description |
Default Value |
| Enable Cross user grouping when using user-task-polling policy. (Disable it if user's task can not merge each other) |
false |
| Description |
Default Value |
| Max pending task per user in scheduler |
1024 |
| Description |
Default Value |
| delegator level zero deletion forward policy, possible option["FilterByBF", "RemoteLoad"] |
FilterByBF |
| Description |
Default Value |
| delegator streaming deletion forward policy, possible option["FilterByBF", "Direct"] |
FilterByBF |
| Description |
Default Value |
| The maximum size of task queue cache in flow graph in query node. |
16 |
| Description |
Default Value |
| Maximum number of tasks executed in parallel in the flowgraph |
1024 |
| Description |
Default Value |
| use partition stats to prune data in search/query on shard delegator |
false |
| Description |
Default Value |
| return min batch size of stream query |
4194304 |
| Description |
Default Value |
| return max batch size of stream query |
134217728 |
| Description |
Default Value |
| parallel factor when to apply pk to bloom filter, default to 4*CPU_CORE_NUM |
4 |
| Description |
Default Value |
| the size for worker querynode client pool |
10 |
| Description |
Default Value |
| TCP/IP address of queryNode. If not specified, use the first unicastable address |
|
| Description |
Default Value |
| TCP port of queryNode |
21123 |
| Description |
Default Value |
| The maximum size of each RPC request that the queryNode can send, unit: byte |
536870912 |
| Description |
Default Value |
| The maximum size of each RPC request that the queryNode can receive, unit: byte |
268435456 |
| Description |
Default Value |
| The maximum size of each RPC request that the clients on queryNode can send, unit: byte |
268435456 |
| Description |
Default Value |
| The maximum size of each RPC request that the clients on queryNode can receive, unit: byte |
536870912 |
