Hi,

I'm looking for a nsqd exporter and came across this project. Awesome!
For now, I just want to say I would recommend to run one exporter instance per nsqd instance and not implement discovery via nsqlookupd because it make it easier to isolate failures to specific instances and leave the instance labeling to prometheus. This is generally considered best practice in the prometheus community.

Was wondering if this project is still being maintained? If not and you'd like a new home for it i'd love to discuss that further.

Hello,

Wnen we trying running nsq_exporter by:
version: '2'
services:
nsqlookupd:
image: nsqio/nsq
command: /nsqlookupd -broadcast-address=nsqlookupd
restart: always
ports:
- "4160:4160"
- "4161:4161"
nsqd:
image: nsqio/nsq
command: /nsqd --lookupd-tcp-address=nsqlookupd:4160 --data-path=/data --broadcast-address=nsqd
restart: always
volumes:
- nsq_data:/data
ports:
- "4150:4150"
- "4151:4151"
nsqadmin:
image: nsqio/nsq
command: /nsqadmin --lookupd-http-address=nsqlookupd:4161
restart: always
ports:
- "4171:4171"
nsq_exporter:
image: lovoo/nsq_exporter:latest
restart: always
command:
- '-nsqd.addr=http://nsqd:4151/stats'
- '-collect=stats.topics,stats.channels,stats.clients'
ports:
- "9117:9117"

volumes:
nsq_data: {}

they not collecnt any data:
go_memstats_alloc_bytes 1.610192e+06
go_memstats_alloc_bytes_total 2.4429984e+07
go_memstats_buck_hash_sys_bytes 1.444735e+06
go_memstats_frees_total 30021
go_memstats_gc_sys_bytes 475136
go_memstats_heap_alloc_bytes 1.610192e+06
go_memstats_heap_idle_bytes 5.808128e+06
go_memstats_heap_inuse_bytes 2.121728e+06
go_memstats_heap_objects 5940
go_memstats_heap_released_bytes 0
go_memstats_heap_sys_bytes 7.929856e+06
go_memstats_last_gc_time_seconds 1.4992581719082916e+09
go_memstats_lookups_total 287
go_memstats_mallocs_total 35961
go_memstats_mcache_inuse_bytes 2400
go_memstats_mcache_sys_bytes 16384
go_memstats_mspan_inuse_bytes 24000
go_memstats_mspan_sys_bytes 32768
go_memstats_next_gc_bytes 4.194304e+06
go_memstats_other_sys_bytes 789881
go_memstats_stack_inuse_bytes 458752
go_memstats_stack_sys_bytes 458752
go_memstats_sys_bytes 1.1147512e+07
http_request_duration_microseconds{handler="prometheus",quantile="0.5"} 1768.584
http_request_duration_microseconds{handler="prometheus",quantile="0.9"} 5565.156
http_request_duration_microseconds{handler="prometheus",quantile="0.99"} 15772.45
http_request_duration_microseconds_sum{handler="prometheus"} 71702.471
http_request_duration_microseconds_count{handler="prometheus"} 26
http_request_size_bytes{handler="prometheus",quantile="0.5"} 276
http_request_size_bytes{handler="prometheus",quantile="0.9"} 276
http_request_size_bytes{handler="prometheus",quantile="0.99"} 276
http_request_size_bytes_sum{handler="prometheus"} 6963
http_request_size_bytes_count{handler="prometheus"} 26
http_requests_total{code="200",handler="prometheus",method="get"} 26
http_response_size_bytes{handler="prometheus",quantile="0.5"} 1447
http_response_size_bytes{handler="prometheus",quantile="0.9"} 1467
http_response_size_bytes{handler="prometheus",quantile="0.99"} 7046
http_response_size_bytes_sum{handler="prometheus"} 42655
http_response_size_bytes_count{handler="prometheus"} 26
nsq_exporter_scrape_duration_seconds{result="success",quantile="0.5"} 0.0007699250000000001
nsq_exporter_scrape_duration_seconds{result="success",quantile="0.9"} 0.001025803
nsq_exporter_scrape_duration_seconds{result="success",quantile="0.99"} 0.014166406000000001
nsq_exporter_scrape_duration_seconds_sum{result="success"} 0.03366974
nsq_exporter_scrape_duration_seconds_count{result="success"} 26
process_cpu_seconds_total 0.09
process_max_fds 65536
process_open_fds 8
process_resident_memory_bytes 1.2742656e+07
process_start_time_seconds 1.49925804813e+09
process_virtual_memory_bytes 1.8702336e+07

How we can fix that?

Thank you.

I feel like the metric names could be made more consistent.
Every gauge exported by nsq_exporter uses _counter as its suffix, except depth and backend_depth.

Any thoughts on updating their names?

When aggregating, it's somewhat jarring to see two metrics with different suffixes being added together:

The expression (depth + in_flight_count) looks strange...
even though they're both gauges.