Add metrics endpoint and request metrics

[amila-ku]

Resolves #3144

This pull request is to add /metrics endpoint and generally used metrics.
It exposes default prometheus metrics and custom metrics for request endpoints.

This PR does not try to cover all metrics to keep it simple. If this looks good. I could add few more that will be useful.

How to test:
once Ollama server is running pull a model and list(or any other Ollama actions)

curl http://127.0.0.1:11434/metrics

example custom metrics(not all are shown since i tried only few commands):

curl http://127.0.0.1:11434/metrics | grep -i ollama
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  6664    0  6664    0     0   519k      0 --:--:-- --:--:-- --:--:--  542k
# HELP ollama_model_list_requests_total The total number of model list requets that have been attempted.
# TYPE ollama_model_list_requests_total counter
ollama_model_list_requests_total{action="list",status="OK",status_code="200"} 1
# HELP ollama_model_pull_requests_total The total number of model pulls that have been attempted.
# TYPE ollama_model_pull_requests_total counter
ollama_model_pull_requests_total{action="pull",status="OK",status_code="200"} 1
# HELP ollama_model_requests_total The total number of requests on all endpoints.
# TYPE ollama_model_requests_total counter
ollama_model_requests_total{action="all",status="OK",status_code="200"} 6

Full output:

# curl http://127.0.0.1:11434/metrics                 
# HELP go_gc_duration_seconds A summary of the pause duration of garbage collection cycles.
# TYPE go_gc_duration_seconds summary
go_gc_duration_seconds{quantile="0"} 1.0458e-05
go_gc_duration_seconds{quantile="0.25"} 5.4042e-05
go_gc_duration_seconds{quantile="0.5"} 6.9792e-05
go_gc_duration_seconds{quantile="0.75"} 0.000121417
go_gc_duration_seconds{quantile="1"} 0.00423525
go_gc_duration_seconds_sum 0.019693795
go_gc_duration_seconds_count 42
# HELP go_goroutines Number of goroutines that currently exist.
# TYPE go_goroutines gauge
go_goroutines 9
# HELP go_info Information about the Go environment.
# TYPE go_info gauge
go_info{version="go1.22.1"} 1
# HELP go_memstats_alloc_bytes Number of bytes allocated and still in use.
# TYPE go_memstats_alloc_bytes gauge
go_memstats_alloc_bytes 2.429592e+06
# HELP go_memstats_alloc_bytes_total Total number of bytes allocated, even if freed.
# TYPE go_memstats_alloc_bytes_total counter
go_memstats_alloc_bytes_total 9.359476e+07
# HELP go_memstats_buck_hash_sys_bytes Number of bytes used by the profiling bucket hash table.
# TYPE go_memstats_buck_hash_sys_bytes gauge
go_memstats_buck_hash_sys_bytes 11319
# HELP go_memstats_frees_total Total number of frees.
# TYPE go_memstats_frees_total counter
go_memstats_frees_total 545225
# HELP go_memstats_gc_sys_bytes Number of bytes used for garbage collection system metadata.
# TYPE go_memstats_gc_sys_bytes gauge
go_memstats_gc_sys_bytes 3.465616e+06
# HELP go_memstats_heap_alloc_bytes Number of heap bytes allocated and still in use.
# TYPE go_memstats_heap_alloc_bytes gauge
go_memstats_heap_alloc_bytes 2.429592e+06
# HELP go_memstats_heap_idle_bytes Number of heap bytes waiting to be used.
# TYPE go_memstats_heap_idle_bytes gauge
go_memstats_heap_idle_bytes 6.660096e+06
# HELP go_memstats_heap_inuse_bytes Number of heap bytes that are in use.
# TYPE go_memstats_heap_inuse_bytes gauge
go_memstats_heap_inuse_bytes 5.136384e+06
# HELP go_memstats_heap_objects Number of allocated objects.
# TYPE go_memstats_heap_objects gauge
go_memstats_heap_objects 12324
# HELP go_memstats_heap_released_bytes Number of heap bytes released to OS.
# TYPE go_memstats_heap_released_bytes gauge
go_memstats_heap_released_bytes 6.201344e+06
# HELP go_memstats_heap_sys_bytes Number of heap bytes obtained from system.
# TYPE go_memstats_heap_sys_bytes gauge
go_memstats_heap_sys_bytes 1.179648e+07
# HELP go_memstats_last_gc_time_seconds Number of seconds since 1970 of last garbage collection.
# TYPE go_memstats_last_gc_time_seconds gauge
go_memstats_last_gc_time_seconds 1.7125328358982313e+09
# HELP go_memstats_lookups_total Total number of pointer lookups.
# TYPE go_memstats_lookups_total counter
go_memstats_lookups_total 0
# HELP go_memstats_mallocs_total Total number of mallocs.
# TYPE go_memstats_mallocs_total counter
go_memstats_mallocs_total 557549
# HELP go_memstats_mcache_inuse_bytes Number of bytes in use by mcache structures.
# TYPE go_memstats_mcache_inuse_bytes gauge
go_memstats_mcache_inuse_bytes 4800
# HELP go_memstats_mcache_sys_bytes Number of bytes used for mcache structures obtained from system.
# TYPE go_memstats_mcache_sys_bytes gauge
go_memstats_mcache_sys_bytes 15600
# HELP go_memstats_mspan_inuse_bytes Number of bytes in use by mspan structures.
# TYPE go_memstats_mspan_inuse_bytes gauge
go_memstats_mspan_inuse_bytes 123840
# HELP go_memstats_mspan_sys_bytes Number of bytes used for mspan structures obtained from system.
# TYPE go_memstats_mspan_sys_bytes gauge
go_memstats_mspan_sys_bytes 179520
# HELP go_memstats_next_gc_bytes Number of heap bytes when next garbage collection will take place.
# TYPE go_memstats_next_gc_bytes gauge
go_memstats_next_gc_bytes 5.554272e+06
# HELP go_memstats_other_sys_bytes Number of bytes used for other system allocations.
# TYPE go_memstats_other_sys_bytes gauge
go_memstats_other_sys_bytes 1.004569e+06
# HELP go_memstats_stack_inuse_bytes Number of bytes in use by the stack allocator.
# TYPE go_memstats_stack_inuse_bytes gauge
go_memstats_stack_inuse_bytes 786432
# HELP go_memstats_stack_sys_bytes Number of bytes obtained from system for stack allocator.
# TYPE go_memstats_stack_sys_bytes gauge
go_memstats_stack_sys_bytes 786432
# HELP go_memstats_sys_bytes Number of bytes obtained from system.
# TYPE go_memstats_sys_bytes gauge
go_memstats_sys_bytes 1.7259536e+07
# HELP go_threads Number of OS threads created.
# TYPE go_threads gauge
go_threads 15
# HELP ollama_model_list_requests_total The total number of model list requets that have been attempted.
# TYPE ollama_model_list_requests_total counter
ollama_model_list_requests_total{action="list",status="OK",status_code="200"} 1
# HELP ollama_model_pull_requests_total The total number of model pulls that have been attempted.
# TYPE ollama_model_pull_requests_total counter
ollama_model_pull_requests_total{action="pull",status="OK",status_code="200"} 1
# HELP ollama_model_requests_total The total number of requests on all endpoints.
# TYPE ollama_model_requests_total counter
ollama_model_requests_total{action="all",status="OK",status_code="200"} 7
# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds.
# TYPE process_cpu_seconds_total counter
process_cpu_seconds_total 73.31
# HELP process_max_fds Maximum number of open file descriptors.
# TYPE process_max_fds gauge
process_max_fds 1.048576e+06
# HELP process_open_fds Number of open file descriptors.
# TYPE process_open_fds gauge
process_open_fds 9
# HELP process_resident_memory_bytes Resident memory size in bytes.
# TYPE process_resident_memory_bytes gauge
process_resident_memory_bytes 2.942976e+08
# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.
# TYPE process_start_time_seconds gauge
process_start_time_seconds 1.71253222787e+09
# HELP process_virtual_memory_bytes Virtual memory size in bytes.
# TYPE process_virtual_memory_bytes gauge
process_virtual_memory_bytes 2.536882176e+09
# HELP process_virtual_memory_max_bytes Maximum amount of virtual memory available in bytes.
# TYPE process_virtual_memory_max_bytes gauge
process_virtual_memory_max_bytes 1.8446744073709552e+19
# HELP promhttp_metric_handler_requests_in_flight Current number of scrapes being served.
# TYPE promhttp_metric_handler_requests_in_flight gauge
promhttp_metric_handler_requests_in_flight 1
# HELP promhttp_metric_handler_requests_total Total number of scrapes by HTTP status code.
# TYPE promhttp_metric_handler_requests_total counter
promhttp_metric_handler_requests_total{code="200"} 3
promhttp_metric_handler_requests_total{code="500"} 0
promhttp_metric_handler_requests_total{code="503"} 0

[amila-ku]

@yuliyantsvetkov @jmorganca It would be great if you can take a look.

[a-mader]

This would be great

[joshcarp]

I think if observability is going to be added it should use the latest OpenTelmetry libraries instead of prometheus_client, especially because they're now stable and there's a lot more support now. prometheus only supports metrics, also, whereas I see it would be very valuable to add traces for observability purposes.

Happy to chat more about it on discord

[amila-ku]

@joshcarp I checked the OpenTelemetry libraries and your suggestion makes sense. It would be better to use Otel libraries. I will update the PR or create new one when i get some free time.

[joshcarp]

There's also a company called traceloop which does LLM observability who are spearheading a lot of the LLM observability stuff in Otel AFAIK. Might be worth seeing what they're doing because a lot of this seems like they're implementing/standardiing

[patcher9]

Hey folks, I am the maintainer for OpenLIT (OTel-native observability for LLMs) and we do have auto instrumentation for Ollama in OpenLIT. It follows the OTel semcov for GenAI that was merged by us all in the OTel community.

We cover metrics and traces from the application side, GPU metrics are something that we haven't yet covered in this though yet much like the other LLM Observability libraries but something we are actively working on. It generally invloves pulling out metrics from nvidia-smi kinda similar to how its done in gpustat library