Skip to content

[KVCache] Add free_cpu_block_num gauge metric #7856

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
liyonghua0910 wants to merge 1 commit into
base: develop
Choose a base branch
from
+15 −0
Open

[KVCache] Add free_cpu_block_num gauge metric #7856

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions docs/online_serving/metrics.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@ After FastDeploy is launched, it supports continuous monitoring of the FastDeplo
| KV Cache | `fastdeploy:free_gpu_block_num` | Gauge | Number of free GPU blocks in cache | count |
| KV Cache | `fastdeploy:max_gpu_block_num` | Gauge | Total number of GPU blocks initialized at startup | count |
| KV Cache | `fastdeploy:max_cpu_block_num` | Gauge | Total number of CPU blocks initialized at startup | count |
| KV Cache | `fastdeploy:free_cpu_block_num` | Gauge | Number of free CPU blocks in cache | count |
| KV Cache | `fastdeploy:available_gpu_resource` | Gauge | Ratio of available GPU blocks to total GPU blocks | % |
| KV Cache | `fastdeploy:gpu_cache_usage_perc` | Gauge | GPU KV cache utilization | % |
| KV Cache | `fastdeploy:send_cache_failed_num` | Counter | Total number of cache send failures | count |
Expand Down
1 change: 1 addition & 0 deletions docs/zh/online_serving/metrics.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@
| KV缓存 | `fastdeploy:free_gpu_block_num` | Gauge | 缓存中的可用块数 | 个 |
| KV缓存 | `fastdeploy:max_gpu_block_num` | Gauge | 服务启动时确定的 GPU 总块数 | 个 |
| KV缓存 | `fastdeploy:max_cpu_block_num` | Gauge | 服务启动时确定的 CPU 总块数 | 个 |
| KV缓存 | `fastdeploy:free_cpu_block_num` | Gauge | 缓存中的可用 CPU 块数 | 个 |
| KV缓存 | `fastdeploy:available_gpu_resource` | Gauge | 可用块占比，即可用 GPU 块数量 / 最大GPU块数量| 百分比 |
| KV缓存 | `fastdeploy:gpu_cache_usage_perc` | Gauge | GPU 上的 KV 缓存使用率 | 百分比 |
| KV缓存 | `fastdeploy:send_cache_failed_num` | Counter | 发送缓存失败的总次数 | 个 |
Expand Down
6 changes: 6 additions & 0 deletions fastdeploy/cache_manager/prefix_cache_manager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,6 +132,7 @@ def __init__(
main_process_metrics.max_cpu_block_num.set(self.num_cpu_blocks)
main_process_metrics.available_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.free_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.free_cpu_block_num.set(self.num_cpu_blocks)
main_process_metrics.available_gpu_resource.set(1.0)

def _get_kv_cache_shape(self, max_block_num):
Expand Down Expand Up @@ -461,6 +462,7 @@ def update_cache_config(self, cache_config):
main_process_metrics.max_cpu_block_num.set(self.num_cpu_blocks)
main_process_metrics.available_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.free_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.free_cpu_block_num.set(self.num_cpu_blocks)
main_process_metrics.available_gpu_resource.set(1.0)

def can_allocate_gpu_blocks(self, num_blocks: int):
Expand Down Expand Up @@ -520,6 +522,7 @@ def allocate_cpu_blocks(self, num_blocks):
logger.info(
f"allocate_cpu_blocks: {allocated_block_ids}, len(self.cpu_free_block_list) {len(self.cpu_free_block_list)}"
)
main_process_metrics.free_cpu_block_num.set(len(self.cpu_free_block_list))
return allocated_block_ids

def recycle_cpu_blocks(self, cpu_block_ids):
Expand All @@ -534,6 +537,7 @@ def recycle_cpu_blocks(self, cpu_block_ids):
heapq.heappush(self.cpu_free_block_list, cpu_block_id)
else:
heapq.heappush(self.cpu_free_block_list, cpu_block_ids)
main_process_metrics.free_cpu_block_num.set(len(self.cpu_free_block_list))

def issue_swap_task(
self,
Expand Down Expand Up @@ -1439,6 +1443,7 @@ def free_cpu_block_ids(self, need_block_num):
logger.info(
"free_cpu_block_ids: after free, " + f"len(self.cpu_free_block_list) {len(self.cpu_free_block_list)}"
)
main_process_metrics.free_cpu_block_num.set(len(self.cpu_free_block_list))
return total_cpu_free_count

def get_block_hash_extra_keys(self, request, start_idx, end_idx, mm_idx):
Expand Down Expand Up @@ -2127,6 +2132,7 @@ def reset(self):
# reset metrics
self.metrics.reset_metrics()
main_process_metrics.free_gpu_block_num.set(len(self.gpu_free_block_list))
main_process_metrics.free_cpu_block_num.set(len(self.cpu_free_block_list))
main_process_metrics.available_gpu_block_num.set(len(self.gpu_free_block_list))
main_process_metrics.available_gpu_resource.set(self.available_gpu_resource)

Expand Down
7 changes: 7 additions & 0 deletions fastdeploy/metrics/metrics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -159,6 +159,7 @@ class MetricsManager:
free_gpu_block_num: "Gauge"
max_gpu_block_num: "Gauge"
max_cpu_block_num: "Gauge"
free_cpu_block_num: "Gauge"
available_gpu_resource: "Gauge"
requests_number: "Counter"
send_cache_failed_num: "Counter"
Expand Down Expand Up @@ -245,6 +246,12 @@ class MetricsManager:
"description": "Number of total CPU blocks determined when service started",
"kwargs": {},
},
"free_cpu_block_num": {
"type": Gauge,
"name": "fastdeploy:free_cpu_block_num",
"description": "Number of free CPU blocks in cache",
"kwargs": {},
},
"available_gpu_resource": {
"type": Gauge,
"name": "fastdeploy:available_gpu_resource",
Expand Down
Loading