io stats per process

[whetherharder]

Feature Request

It would be quite useful to have io statistics (iops, reads/per-sec,writes/sec etc), for paticular process. It would be perfect, to have also network stats per process.
If there is already a way to get such statistics, using existing set of plugins, let me know please

[sparrc]

there's the procstat plugin: https://github.com/influxdata/telegraf/tree/master/plugins/inputs/procstat

[whetherharder]

Sorry for bothering you with io stats they are already present as it is writen in docs!. My problem is that not all metrics were pushed to influxdb

So
select * from "procstat" limit 1
returns only these fields:
cpu_time_guest cpu_time_guest_nice cpu_time_idle cpu_time_iowait cpu_time_irq cpu_time_nice cpu_time_soft_irq cpu_time_steal cpu_time_stolen cpu_time_system cpu_time_user cpu_usage host involuntary_context_switches memory_rss memory_swap memory_vms num_threads pattern pid process_name service_name voluntary_context_switches

while telegraf -test -config input.procstat.custom.zookeeper.conf returns
[root@SRV2-VTG-VM47 telegraf.d]# telegraf -test -config input.procstat.custom.zookeeper.conf

• Plugin: procstat, Collection 1

  procstat,host=SRV2-VTG-VM47,pattern=zookeeper,pid=16022,process_name=telegraf,service_name=zookeeper cpu_time_guest=0,cpu_time_guest_nice=0,cpu_time_idle=0,cpu_time_iowait=0,cpu_time_irq=0,cpu_time_nice=0,cpu_time_soft_irq=0,cpu_time_steal=0,cpu_time_stolen=0,cpu_time_system=0.01,cpu_time_user=0.02,involuntary_context_switches=1i,memory_rss=15986688i,memory_swap=0i,memory_vms=511963136i,num_fds=5i,num_threads=14i,read_bytes=0i,read_count=906i,voluntary_context_switches=53i,write_bytes=3i,write_count=3i 1464736322487619429

My configuration is:
Telegraf - version 0.13.1
InfluxDb - version 0.9.3

May be InfluxDb is too old?

[whetherharder]

Looks like a bug or some misconfiguration
I tried to reproduce the same scenario on influxdb 13.0 and the result was the same -- a part of fields are not pushed to influxdb

I have a guess, that the reason is in some kind of configured or hardcoded limit of fields to transmit. The set fields that is pushed is always identical

[sparrc]

Only possiblity I can think of would be that it's a permissions issue, could you try running sudo -u telegraf telegraf -test -config input.procstat.custom.zookeeper.conf and see if the output is different?

[whetherharder]

I am running telegraf under root as as a service
31.05.2016 18:30 пользователь "Cameron Sparr" notifications@github.com
написал:

there's the procstat plugin:
https://github.com/influxdata/telegraf/tree/master/plugins/inputs/procstat

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#1300 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AD_WapALQ72P7hLk653Yv0OkgB1hCXB9ks5qHFQBgaJpZM4Iqp69
.

[whetherharder]

But i will check this point. Thanks
01.06.2016 16:51 пользователь "Cameron Sparr" notifications@github.com
написал:

Only possiblity I can think of would be that it's a permissions issue,
could you try running sudo -u telegraf telegraf -test -config
input.procstat.custom.zookeeper.conf and see if the output is different?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#1300 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AD_WauY95EyXPSrhs-ECE6Vvv3Nl5Ba2ks5qHY5YgaJpZM4Iqp69
.

[sparrc]

Unless you're rolling your own init.sh or telegraf.service scripts I think it will run under the telegraf user anyways

[sparrc]

@whetherharder any updates?

[Finistere]

I just tried running sudo -u telegraf telegraf -test -config custom_procstat.conf and got:

• Plugin: procstat, Collection 2

  procstat,host=hadoop-node-1,pattern=bash,pid=2129,process_name=bash cpu_time_guest=0,cpu_time_guest_nice=0,cpu_time_idle=0,cpu_time_iowait=0,cpu_time_irq=0,cpu_time_nice=0,cpu_time_soft_irq=0,cpu_time_steal=0,cpu_time_stolen=0,cpu_time_system=0.06,cpu_time_user=0.01,involuntary_context_switches=5i,memory_rss=2183168i,memory_swap=0i,memory_vms=118153216i,num_threads=1i,voluntary_context_switches=441i 1468763178088860035

I don't see any I/O stat in it.

Configuration:
Telegraf 0.13.1
CentOS 7.2

Is there something I might have missed ?

[sparrc]

@Finistere I believe the issue would be that your telegraf user doesn't have permission to gather these stats. What happens when you run as root?

[Finistere]

@sparrc Indeed, with root it works fine. The records are also correctly stored in InfluxDB (version 0.13.0).
Thanks for the help!

[sparrc]

closing as it appears to be a permissions issue.

If there is a specific workaround that should be pointed out, please submit a PR to change the plugin README, thanks

[mossad-zika]

I don't think it works nowadays ..