search

chrisss404 / check-mk-arm

Checkmk for Raspberry Pi.
183 stars 22 forks source link

No services in host view using check-mk-raw-1.5.0p4_0 #6

Closed onexp closed 5 years ago

onexp commented 5 years ago

Hi,

I just installed your latest version check-mk-raw-1.5.0p4_0.stretch_armhf.deb on my Raspberry pi 2. As OS I am using Raspbian Strech Lite.

So far the installation went fine, sites are running and I also installed the agent on some test servers running Ubuntu 16.04. All services were detected with auto discovery and added.

But something is wrong with the service views. If I go to the "All host" section I can see my servers and the overview, how many services are OK, WARN, etc. But if I click on a host, the service view is emtpy. Only for the localhost the view is working. There I can see all services.
But if I go to Main Overview and click under "Service Statistics" on OK, or Total, I get a complete overview of my services, sort by host (same if I go to "All services". I was not able to find the problem and installed your version check-mk-raw-1.4.0p35_0.stretch_armhf.deb This version is running fine and all views are working as expected.

chrisss404 commented 5 years ago

Hi,

I currently monitor 4 hosts and the service view works as expected for all of them.

Things you can try:

My setup

$ lsb_release -a
No LSB modules are available.
Distributor ID: Raspbian
Description:    Raspbian GNU/Linux 9.4 (stretch)
Release:    9.4
Codename:   stretch

$ omd version
OMD - Open Monitoring Distribution Version 1.5.0p4.cre
onexp commented 5 years ago

Thanks for your feedback. Update and fresh install didn't help. I even took another Pi 2 and installed everything from scratch. But I am still not able to see the services, when I go to "All hosts" and click on one host. Then I installed mk_check on another x86 VM Ubuntu 16.04 and I was able to see the services with the new version. So I really don't know, where is the problem. Should be no difference between Pi 3 (you are running) and Pi 2 ?!

chrisss404 commented 5 years ago

I don't think that the issue is related to the Pi version.

I tried to find the difference between the views you described. If you check the links for those views, you have:

All services http://your-pi-host/check_mk/view.py?view_name=allservices

All hosts > one hosts http://your-pi-host/check_mk/view.py?host=rpi&site=some-site&view_name=host

You can see, that the same view.py script is used to render the views. Therefore I can not think of a reason why the host specific view should not work.

However what I have noticed, if you use a non-existing host for the host parameter the view is empty as you described, e.g.:

http://your-pi-host/check_mk/view.py?host=this-host-does-not-exists&site=some-site&view_name=host

Hope that helps, otherwise I have no idea why this is happening to you.

gituser-rk commented 5 years ago

Hello, same strange behavior here (Raspi 3B). The view of one host shows no services, "view.py?view_name=allservices" is showing the services of this host. I have deleted this host and recreated it. The services are shown during the service scan when creating the host in WATO. But later the same - no services in host service view. In my case this host is a "Fritzbox" router. The services of all other hosts show up fine in the individual host service view. As long as I remember this is since version 1.5.0p2 up to 1.5.0p4. Version 1.4.3x was fine.

chrisss404 commented 5 years ago

Hi,

one last try. The only difference to the upstream build is that python is built without --enable-shared flag, as can be seen in python-Makefile.patch. There are suggestions to do this in order to improve performance. However, if this introduces issues it should be removed.

Here is a build without applying the mentioned patch. Hopefully this will fix the issue.

gituser-rk commented 5 years ago

Hello,

tried the "special" build and it made no difference. Just installed the .deb file over and started the OMD instance. But nevertheless thanks very much for your effort

mwiedemann commented 5 years ago

I have exactly the same problem. RPi3 with a fresh install. If I have created a host an added some services the service view is empty: Linkname: http://cmk/default/check_mk/view.py?host=myHost&site=default&view_name=host#

chrisss404 commented 5 years ago

Some more input that hopefully helps narrow down the issue.

Logs

Have you checked the logs of your site? Maybe there is a hint what is causing the issue.

/opt/omd/sites/<your-site>/var/log/apache/error_log
/opt/omd/sites/<your-site>/var/log/apache/access_log
/opt/omd/sites/<your-site>/var/log/web.log

Hostnames

Maybe the issue is related to the used hostname. Here is an overview of my hosts / hostnames:

Hostname Url Services shown
htz1 /check_mk/view.py?host=htz1&site=site&view_name=host Yes
rpi /check_mk/view.py?host=rpi&site=site&view_name=host Yes
up /check_mk/view.py?host=up&site=site&view_name=host Yes
wt /check_mk/view.py?host=wt&site=site&view_name=host Yes

Dependencies

Maybe I have installed a package that is not defined as dependency in the .deb package.

Here is a list of installed packages of my raspberry pi:

dpkg --get-selections | grep -v deinstall > packages.txt
gituser-rk commented 5 years ago

renamed the host from "fritz.box" to "fritzbox" (and setting a IP Address to make sure that the host is still reachable) and voilà: the services are back in host services view!

msperl commented 5 years ago

I have experienced the same thing - hosts with dots are not showing up propperly. similarly the "action menu" does not work with hostnames with dots. I have even added the same host twice - once without domainname (plus ipv4 set) once the fqhn (same effect - with or without ipv4 set).

But everything works properly with 1.5.0p7 on an x86_64 machine (at least the dockerized version from checkmk).

msperl commented 5 years ago

just renaming the hosts changes the behaviour!

Maybe a python library thingy?

msperl commented 5 years ago

So I have started tracing it and it seems to be related to the response of live-status:

OMD[cmk]:~$ cat /tmp/test.sh
#! /bin/sh

for x in kvm kvm.intern; do
    echo "Processing: $x"
    echo "================="
    echo -n "GET hosts\nColumns: host_custom_variable_names host_action_url_expanded host_filename host_check_command host_active_checks_enabled host_perf_data host_pnpgraph_present host_custom_variable_values host_name host_check_type host_address host_notes_url_expanded\nFilter: host_name = $x\nLocaltime: 1540640697\nOutputFormat: python\nKeepAlive: on\nResponseHeader: fixed16\nColumnHeaders: off\n\n" \
    | nc.openbsd -w 1 -U /omd/sites/cmk/tmp/run/live
done

    echo "list hosts"
    echo "================="
    echo -n "GET hosts\nColumns: host_name host_check_type host_address\nLocaltime: 1540640697\nOutputFormat: python\nKeepAlive: on\nResponseHeader: fixed16\nColumnHeaders: off\n\n" \
| nc.openbsd -w 1 -U /omd/sites/cmk/tmp/run/live
OMD[cmk]:~$ sh /tmp/test.sh
Processing: kvm
=================
200         506
[[[u"ADDRESS_4",u"ADDRESS_6",u"TAGS",u"FILENAME",u"ADDRESS_FAMILY"],u"/cmk/pnp4nagios/index.php/graph?host=kvm&srv=_HOST_' class='tips' rel='/cmk/pnp4nagios/index.php/popup?host=kvm&srv=_HOST_",u"/wato/hosts.mk",u"check-mk-host-ping!-w 200.00,80.00% -c 500.00,100.00%",1,u"rta=0.586ms;200.000;500.000;0; pl=0%;80;100;; rtmax=0.629ms;;;; rtmin=0.541ms;;;;",1,[u"10.10.10.250",u"",u"/wato/ cmk-agent ip-v4 ip-v4-only lan no-snmp prod site:cmk tcp wato",u"/wato/hosts.mk",u"4"],u"kvm",0,u"10.10.10.250",u""]]
Processing: kvm.intern
=================
200           3
[]
list hosts
=================
200          64
[[u"kvm",0,u"10.10.10.250"],
[u"kvm.intern",0,u"10.10.10.250"]]

So it seems to be related to live-status!

On the intel instance I get:

Processing: kvm
=================
200         506
[[[u"ADDRESS_4",u"ADDRESS_6",u"TAGS",u"FILENAME",u"ADDRESS_FAMILY"],u"/cmk/pnp4nagios/index.php/graph?host=kvm&srv=_HOST_' class='tips' rel='/cmk/pnp4nagios/index.php/popup?host=kvm&srv=_HOST_",u"/wato/hosts.mk",u"check-mk-host-ping!-w 200.00,80.00% -c 500.00,100.00%",1,u"rta=0.316ms;200.000;500.000;0; pl=0%;80;100;; rtmax=0.495ms;;;; rtmin=0.167ms;;;;",1,[u"10.10.10.250",u"",u"/wato/ cmk-agent ip-v4 ip-v4-only lan no-snmp prod site:cmk tcp wato",u"/wato/hosts.mk",u"4"],u"kvm",0,u"10.10.10.250",u""]]
Processing: kvm.intern
=================
200         527
[[[u"ADDRESS_4",u"ADDRESS_6",u"TAGS",u"FILENAME",u"ADDRESS_FAMILY"],u"/cmk/pnp4nagios/index.php/graph?host=kvm.intern&srv=_HOST_' class='tips' rel='/cmk/pnp4nagios/index.php/popup?host=kvm.intern&srv=_HOST_",u"/wato/hosts.mk",u"check-mk-host-ping!-w 200.00,80.00% -c 500.00,100.00%",1,u"rta=0.718ms;200.000;500.000;0; pl=0%;80;100;; rtmax=0.769ms;;;; rtmin=0.648ms;;;;",1,[u"10.10.10.250",u"",u"/wato/ cmk-agent ip-v4 ip-v4-only lan no-snmp prod site:cmk tcp wato",u"/wato/hosts.mk",u"4"],u"kvm.intern",0,u"10.10.10.250",u""]]
list hosts
=================
200          64
[[u"kvm",0,u"10.10.10.250"],
[u"kvm.intern",0,u"10.10.10.250"]]

So here livestatus reports information for both versions...

msperl commented 5 years ago

a "slow" compile of gcc 7.3.0 shows that it should be sufficient for mk_livestatus. Maybe that would be sufficient... I will try to compile check_mk with that to see if that solves the issue

msperl commented 5 years ago

So my trial was successfull - compiling with 7.3 and the "." problem is gone...

msperl commented 5 years ago

For those of you wanting to give it a try to see if it solves the issue here my deb package: https://github.com/msperl/check-mk-arm/releases/tag/1.5.0p7-gcc7.3-test

chrisss404 commented 5 years ago

@msperl thanks for your effort.

Can you please share your steps how to resolve this issue. I don't know what is meant by a "slow" compile of gcc 7.3.0, can you explain it? Did you compile gcc 7.3.0 from source or did you use a pre-compiled binary, if so which one?

msperl commented 5 years ago

I have just compiled gcc the way our original source gcc was compiled (just replaced the version for all practical purposes)

I am actually compiling the whole as a series of docker containers - this should become available in my clone of the repo soon.

msperl commented 5 years ago

here the initial version that builds check_mk via a series of docker images https://github.com/msperl/check-mk-arm/tree/dockerize

I am thinking that maybe the default docker images of GCC would also be valid so that we do not have to compile gcc on our own... (but that may leave out the rpi 1 and 2's - to get tested...)

I am looking into that as soon as I have finished building a docker image for running cmk based on the Dockerfile from check_mk itself...

msperl commented 5 years ago

the above branch now also contains the dockerized version of check_mk using the same basic image Dockerfile that checkmk provides with its Sourcecode (besides the removed gpg verification of the debian package, which the build processs currently does not support...).

So you can run:

docker container run -dit -p 8080:5000 -v /omd/sites --name check_mk -v /etc/localtime:/etc/localtime --restart always msperl/check_mk:1.5.0p7-7.3.0-arm

to test the image inside a docker environment on arm - more information on the image can be found here: https://hub.docker.com/r/checkmk/check-mk-raw/ for details...

chrisss404 commented 5 years ago

Should be fixed with 1.5.0p8.