Items keep going to unsupported

[mgapinski]

Hi,

I have a problem which makes me want to pull out my hair in despair...

I am monitoring MySQL performance with a script that pulls all the information by running a set of show * command on the server every minute and puts all the output into cache file (it's located in shm for better performance). Collecting the data is done by a user parameter like (/etc/zabbix/cache is linked to /dev/shm/cache):

UserParameter=mysql.getparam[*], egrep "^$1:" /etc/zabbix/cache/check_mysql.cache | awk -F':' '{ print $$2; }'

Everything goes smoothly until (without any apparent reason) items become unsupported. Zabbix complains that:

2036:20111031:152213.850 item [HOSTNAME:mysql.getparam[Threads_created]] became not supported: Received value [] is not suitable for value type [Numeric (unsigned)] and data type [Decimal]

BUT when I run that same parameter with zabbix_get:

root@ZBX_SVR:~# zabbix_get -s HOSTNAME -k "mysql.getparam[Threads_created]"
67

I am perplexed and on the verge of going insane because of that behavior...
Sometimes it will stay unsupported for a few hours, and sometimes only 15 or so minutes. Does someone have an idea why is it happening?

I would be grateful for any advice

P.S.
I am running latest trunk server and agent

[Colttt]

hello,

how long run the query?? i think you must set the timeout to 30 in the zabbix-config-files.

[mgapinski]

It takes under 1 second to run the polling script, and fetching single values is done in about 0.1 second. Based on that I don't think it times out.

[mward29]

same problem

I'm having the same problem. was a fix ever found for this?

[[email protected]]

File rewrite

Hep

I would suggest that you set the recheck unsupported vaules to 60 sec inorder to quickly regain the data.

Im guessing that you are running a script on the server to populate the data into the file ... my guess is that some times when the zabbix agent pulls the data the other script is rewriting the file causing it to give bogus data.

If my hunch is correct this is a common issue with all sorts of data pulling.

I would suggest that the script on the server dont populate the file right away, but make a temp file and then move it to the other file ... depending on the file system this will fix the issue or make it very rare.

/lbl

[sean]

I too lets lots of flapping between the supported and unsupported state.

item [myhost.ch:squid.mean_obj_size] became not supported: Received value [] is not suitable for value type [Numeric (float)]
..
item [myhost.ch:squid.mean_obj_size] became supported

It is happening for many apache, mysql and squid items all of which call an external script.

zabbixsend workd fine, runnign the scripts ono the command line with "time" show that scripts are very fast (10-100ms).

It really does look like a timeout on the serverside, and its a difference noticed having migrated from zabbix 1.8 to 2.0.
Although the server should wait upto 30 secs, it looks more like max 50ms or so to me..

[[email protected]]

Check the checks.

Zabbix clearly states that the data it got when it pulled the item was "[]" empty.

Im guessing if you do a:

jot 100 0 | while read a ; do zabbix_get -s myhost.ch -k squid.mean_obj_size ; done

You will see a few empty lines because the check dident return a value.

/lbl

[sean]

Many lines with 18.30 some back, and two with ZBX_NOTSUPPORTED. It takes 16 secs in total and there is a short wait when the NOTSUPPORTED lines show.

Switching on debug level 4 on the v1.8.11 client shows a bit more detail about the entries that work, and those that don't :

22789:20130409:144654.845 Run remote command [squidclient -p80 mgr:info|grep 'Mean Object Size:'|cut -d':' -f2|awk '{print $1}'] Result [5] [18.30]...
22789:20130409:144654.845 Sending back [18.30]

22790:20130409:144654.853 Processing request.
22790:20130409:144654.853 Requested [squid.mean_obj_size]
22790:20130409:144654.853 In zbx_popen() command:'squidclient -p80 mgr:info|grep 'Mean Object Size:'|cut -d':' -f2|awk '{print $1}''
22790:20130409:144654.853 End of zbx_popen():7
23712:20130409:144654.854 zbx_popen(): executing script
22788:20130409:144655.749 In update_cpustats()
22788:20130409:144655.749 End of update_cpustats()
22788:20130409:144656.749 In update_cpustats()
22788:20130409:144656.750 End of update_cpustats()
22788:20130409:144657.750 In update_cpustats()
22788:20130409:144657.752 End of update_cpustats()
22790:20130409:144657.853 In zbx_waitpid()
22790:20130409:144657.853 zbx_waitpid() killed by signal 15
22790:20130409:144657.854 End of zbx_waitpid():23712
22790:20130409:144657.854 Sending back [ZBX_NOTSUPPORTED]
22789:20130409:144657.862 Processing request.
22789:20130409:144657.862 Requested [squid.mean_obj_size]
22789:20130409:144657.862 In zbx_popen() command:'squidclient -p80 mgr:info|grep 'Mean Object Size:'|cut -d':' -f2|awk '{print $1}''
22789:20130409:144657.863 End of zbx_popen():7
23717:20130409:144657.863 zbx_popen(): executing script

Does it not look like the external command is conflicting with update_cpustats() for resources and losing?
Any suggestions on how to interpret the above? Thanks in advance..

EDIT: Upgraded the client to v2.0.5. The above test now takes 36 secs and NOTSUPPORTED come up about 3 in every 15 connections, so not an improvement :-(

[elyroha]

Value should be a JSON object

Buen dia
Tengo demasiadas dudas con Zabbix, y poco a poco he ido aclarando, sin embargo, con los servidores Linux, me esta apareciendo el siguiente mensaje Value should be a JSON object
Necesito ayuda con esto.... gracias

[caotic]

Try to switch the item from "numeric unsigned" to "numeric floating"

[elyroha]

Hi!intento colocar los triggers ara los servidores linux, y la mayoria del mismo dice que no es soportado. Quisiera conocer que tipo de triggers puedo realizar para los servidores Linux, porque ni un simple ping, puedo realizar