Ideas on how best to monitor Qmail stats

[eWebtricity]

Any ideas on ways to monitor Qmail stats?

[sauron]

Use external params in zabbix agent. If scripts works slowly (run time more 0.5 secs) use zabbix_trapper

[eWebtricity]

My graphs are staying at zero becuase I think Zabbix is reading the last line of output, which is a zero. Any ideas how to pull each line of stats individually, i can't seem to make it happen yet.

I'm using this command line argument to base my graphs stats from

[root@server zabbix]# /var/qmail/bin/qmail-qstat
messages in queue: 40
messages in queue but not yet preprocessed: 0

Here's what I'm trying to use

[root@titan zabbix]# su - root -c /var/qmail/bin/qmail-qstat |cut -f2 -d":"
 40
 0

I think it's staying at zero because of the last line which is always zero.

[Alexei]

/var/qmail/bin/qmail-qstat |cut -f2 -d":"|head -1 is probably what you want.

[eWebtricity]

That works great! Thanks.

I've added this line to the server zabbix_agentd.conf

#
# Added to check the number of emails in queue
#
UserParameter=qmail[qmailstat],su - root -c "/var/qmail/bin/qmail-qstat |cut -f2 -d":"|head -1"

and then I setup an item in the mgmt gui with these parameters

Description: Messages in Queue
Host: hostname
Type: Zabbix Agent
Key: qmail[qmailstat]
Units: <empty>
Multiplier: <empty>
Update Interval: 60
Keep History: 7
Status: Monitored
Type Information: Numeric
Store Value: Delta

But I'm getting this error in the zabbix_agentd.log on the server i'm monitoring and the item in the mgmt gui changes to "Not Supported" so it's failing for some reason.

[root@server zabbix]# tail -f /tmp/zabbix_agentd.log
027700:20050506:093107 Got signal. Exiting ...
027701:20050506:093107 Got signal. Exiting ...
027699:20050506:093107 Got signal. Exiting ...
027697:20050506:093107 Got signal. Exiting ...
028065:20050506:093113 zabbix_agentd started
028066:20050506:093113 zabbix_agentd 28066 started
028067:20050506:093113 zabbix_agentd 28067 started
028068:20050506:093113 zabbix_agentd 28068 started
028069:20050506:093113 zabbix_agentd 28069 started
028070:20050506:093113 zabbix_agentd 28070 started
standard in must be a tty

[eWebtricity]

Looks like it's a permissions problem. Zabbix doesn't like it when i try to su to the qmail user and it complains when zabbix can't access the qmail directories. I'm not sure how to get around this. Latest error message

[root@server zabbix]# find: queue/mess/*: Permission denied

[eWebtricity]

Using this External Parameter in the zabbix_agentd.conf

UserParameter=qmail[qmailstat],/var/qmail/bin/zabbix-qmail

Getting this error message on the zabbix_agentd.log

005314:20050510:190834 zabbix_agentd 5314 started
005315:20050510:190834 zabbix_agentd 5315 started
find: queue/mess/*: Permission denied
find: queue/mess/*: Permission denied

Zabbix doesn't seem to like the use of sudo or su to execute the qmail-qstat script. Any suggestions?

[johnl]

Use cron job to output result to a file

what you could do is to create a cron job that runs this script and output the result to a file that zabbix has permissions to read. we do this for a number of services we monitor. it works but at the moment we are experiencing a problem whereby the UserParameter switches to Unsupported after a period of time. let me know if you experience this too.

regards
john

[Alexei]

it works but at the moment we are experiencing a problem whereby the UserParameter switches to Unsupported after a period of time. let me know if you experience this too.

I believe it happens because you're doing something like this:

1. command >file
2. UserParameter=qmail, cat file

This is wrong, because when 1. is executed the file becomes (for a 0.01 sec) empty, therefore ZABBIX sets status of the item to Unsupported.

Do this instead:

1. command >file.tmp
2. mv file.tmp file; rm file.tmp
3. UserParameter=qmail, cat file

[johnl]

thanks alexei,

you were exactly right. outputting the result to a tmp file and then moving it so zabbix can access it has done the trick.

thanks for the help and the great software.

john

[xaeth]

I don't know if this would be beneficial to anyone, and I haven't run it for very long, but here is how I'm doing this. I realize it might be a bit much, but I was trying to keep the code in the zabbix_agentd.conf 1 liners cleaner, and still get all 3 portions of the queue. The only thing I am trying to resolve is right this second I am seeing a bit of a discrepancy between qmailctl stat and qmqtool -s, but i'm not sure where it originates from. If I cant figure it out I may get add qmailctl stat's count just for the sake of finding out the nature of the discrepancy.

To view the stats I use qmqtool (which can be found at jeremy.kister.net), which is run by the following simple script as a cronjob (I do it every minute, but i'm a touch more paranoid for specific reasons). Below are the files, or excerpt

qmailstats.sh

#!/bin/sh
/use/local/qmqtool/qmqtool -s > /tmp/queue.tmp
head -1 /tmp/queue.tmp | cut -f2 -d":" > /tmp/queue.local
head -2 /tmp/queue.tmp | tail -1 | cut -f2 -d":" > /tmp/queue.remote
tail -1 /tmp/queue.tmp | cut -f2 -d":" > /tmp/queue.todo

and from zabbix_agentd.conf

### Monitoring qmail queue
### Accompanies /etc/cron.etc/qmailqueue
UserParameter=qmail[qmaillocal],cat /tmp/queue.local
UserParameter=qmail[qmailremote],cat /tmp/queue.remote
UserParameter=qmail[qmailtodo],cat /tmp/queue.todo

and add then add to crontab something like: (change the 1 to 5 for 5 minute increments.

*/1 * * * * root /etc/cron.etc/qmailqueue

Inside the web interface I did the following (i like keeping data extended periods):

Description: Messages in qmail <whichever> queue
Host: <box>
Type: Zabbix agent
Key: qmail[qmail<whichever>]
Units: <none>
Use multiplier: <none>
Update interval (in sec): 60
Keep history (in days): 90
Keep trends (in days): 365
Status: Monitored
Type of information: Numeric
Store value: As is

[shalako]

I'm trying to set up qmail queue monitoring like this, but am not getting any data.

zabbix_agentd.log:

025050:20070207:142008 Can't add user specifed key "qmail[qmailstat]". Incorrect key!
025052:20070207:142008 zabbix_agentd started. ZABBIX 1.1.
025053:20070207:142008 zabbix_agentd 25053 started
025055:20070207:142008 zabbix_agentd 25055 started
025056:20070207:142008 zabbix_agentd 25056 started
025057:20070207:142008 zabbix_agentd 25057 started
025058:20070207:142008 zabbix_agentd 25058 started
025055:20070207:142111 Timeout while answering request
025053:20070207:142608 Timeout while answering request
025053:20070207:142809 Timeout while answering request

zabbix_agentd.conf:

#
# Added to check the number of emails in queue
#
UserParameter=qmail[qmailstat],cat /home/zabbix/qmailstat

qmailstat.sh (cronjob every 10 min):

#! /bin/bash
/var/qmail/bin/qmail-qstat |cut -f2 -d":"|head -1 > /home/zabbix/qmailstat.tmp
cp /home/zabbix/qmailstat.tmp /home/zabbix/qmailstat
chown zabbix:zabbix /home/zabbix/qmailstat
rm /home/zabbix/qmailstat.tmp

/home/zabbix/qmailstat:

169

zabbix gui item config:

Description 	Qmail Messages In Queue
Type 	ZABBIX agent
Key 	qmail[qmailstat]
Type of information 	Numeric (64bit integer)
Units 	blank
Use multiplier 	Do Not Use
Update interval (in sec) 	60
Keep history (in days) 	7
Keep trends (in days) 	365
Status 	Active
Store value 	As is
Show value As is

After a short time, Status for this item turns to Not Supported. Why?

[shalako]

from zabbix_server.log

002930:20070207:151600 Parameter [qmail[qmailstat]] is not supported by agent on host [tin01.2vega.com] Old status [0]

[shalako]

Figured it out. Server and agent don't like the 'qmailstat' in brackets for UserParameter and Key, as in:

UserParameter=qmail[qmailstat],cat /home/zabbix/qmailstat

this worked:

UserParameter=qmail,cat /home/zabbix/qmailstat

I updated the key in the gui.

[marcelein]

UserParameter=qmail.qmaillocal,cat /tmp/queue.local
UserParameter=qmail.qmailremote,cat /tmp/queue.remote
UserParameter=qmail.qmailtodo,cat /tmp/queue.todo
UserParameter=qmail.procs,ps aux | grep qmail | wc -l

my mailserver screen has x columns for x servers
and 3 rows each showing a custom queue graph with local/remote/todo items
a graph with running qmail processes (qmail.procs) and on the last row there is the output of a transaction script which sends (authentificated) email to a native autoresponder and checks if this email comes back (script availible here)
so i can see every problem related on the timeline to fix issues

a nice feature would be to grep qmail's different processes like qmailr, qmaild, qmaill, qmails (look at the qmail manual) and post them on a zabbix graph