jabbix - enterprise - joke ?

[highland]

Hello

I've been testing jabbix 1.8 for the last few days.
I've tried if it's possible to use it as enterprise management system for ~500 network devices and ~2k servers.
And i am very disappointed.

1. Architecture - it's a fail. There is only one level of templates (even in simple cacti there are three levels). What are the effects ? some examples:
a) i want to add 5 cisco routers (for example 7603): i need to create five different teplates for each router, in each template i need to add Items, in each item i need to type snmp OID, than i have to name each Item differently (because each interface on each router have different names!). All of this must be done manually. Even in cacti this is automated (cacti grabs snmp oids and propose interfaces traffic which could be graphed).
b) what if i have 100 identical cisco switches and i want to change color of incoming traffic from red to blue ? I need to manually change all 100 templates, and all items in each templates !
c) what if i have 100 identical cisco switches and i have changed snmp password on them ? I need to manually change all 100 templates and all items in each templates ! (i could use mass update but still it's a great work to do).
d) what if i want to display interface statistics of my routers in Mbit/s not int MB/s ? I can not.
e) what if i want to change legend on the graph ? i can not. I have to create other Item and add it to the graph.
f) what if i want to change name of Item (displayed on the graph) ? I can not - i have to create new Item.
g) i've created one item in my new template. I want to add 100 similar items (interface traffic for another 100 interfaces) - i can not. There is no clone options - i have to add it manually typing all 100 Items with all parameters like timers, units, history...
h) cisco templates from: http://www.zabbix.com/wiki/templates/start - none of them works (no error durring import, no errors in logs but the template does not appear after import). Existing templates for cisco: 87x, 2960, old pix 515/525 - it's definitely not enterprise gear.

Of course i found some other bugs, but it's not the issue here.
The summary for zabbix is: it's not for enterprise, and it's good only for monitoring PC stations using zabbix agent - definitely not for network gear, and other SNMP monitored gear. Moreover zabbix archtecture is not flexible and is much worse then other opensource available on the market.

I really appreciate your work, i know it's very difficult to spare your free time to build such opensource. But let's face a truth. Am i wrong ?

Best regards,
Michal

[nelsonab]

Hello

I've been testing jabbix 1.8 for the last few days.
I've tried if it's possible to use it as enterprise management system for ~500 network devices and ~2k servers.
And i am very disappointed.

Sorry to hear you're disappointed.

1. Architecture - it's a fail. There is only one level of templates (even in simple cacti there are three levels). What are the effects ? some

That's a pretty strong statement and I'm sure there would be many of us who would willingly go toe to toe with you on that. I think Nagios/Cacti is an epic fail, but that's just me. At a former employer I saw a guide that said "31 easy steps to add cacti graphs."

examples:
a) i want to add 5 cisco routers (for example 7603): i need to create five different teplates for each router, in each template i need to add Items, in each item i need to type snmp OID, than i have to name each Item differently (because each interface on each router have different names!). All of this must be done manually. Even in cacti this is automated (cacti grabs snmp oids and propose interfaces traffic which could be graphed).

Nobody is going to argue with you on the frustrations of SNMP, but I think there is a feature you have not tried.
http://www.zabbix.com/documentation/.../advanced_snmp
Now to make it work as it's shown you need to have net_snmp set up correctly with the proper MIB's, but you can use that to do dynamic table lookups.

Also why would you create 5 different templates for the same type of router? Why not create one master template, and some sub templates you can mix in based on function. A host can link to multiple templates, and templates can even link to more templates. Also you will find that Cisco uses common OID's so breaking out your needs into sub-templates might fit your needs more closely than you realize.

b) what if i have 100 identical cisco switches and i want to change color of incoming traffic from red to blue ? I need to manually change all 100 templates, and all items in each templates !

Again, why would you have 100 templates for 100 devices? Are they all that unique?

As for how to change the color of your graphs, you have three choices
1) Use the Web GUI
2) Use the API
3) Directly manipulate the DB (The fastest, but not for newcomers)

c) what if i have 100 identical cisco switches and i have changed snmp password on them ? I need to manually change all 100 templates and all items in each templates ! (i could use mass update but still it's a great work to do).

What about using user defined macros for the SNMP community/username?

Line 586 in src/zabbix_server/poller/poller.c (1.8.4) does just that if you are curious. (Sorry I've been spending a lot of time deep in the code).

For more information about Macros, including usermacros you can go here:
http://www.zabbix.com/documentation/...onfig/macros?s[]=macros

d) what if i want to display interface statistics of my routers in Mbit/s not int MB/s ? I can not.

It's called units, you can edit them on a per item basis, or select multiple items and then click on mass update. Zabbix understands the difference between B and b, but if that does not work you can also use the custom multiplier.

e) what if i want to change legend on the graph ? i can not. I have to create other Item and add it to the graph.

I'm sorry I'm not quite following you on this. If you are referring to the basic graph, the title comes from the items description. Otherwise if you want to graph multiple things together on one graph, yes you can then create a graph, and you can change it's title. There are some edge cases however where it would be nicer to have a bit more granularity, but for the most part they are edge cases.

f) what if i want to change name of Item (displayed on the graph) ? I can not - i have to create new Item.

Again not quite sure where you're coming from on this. Why can you not edit a pre-existing item?

g) i've created one item in my new template. I want to add 100 similar items (interface traffic for another 100 interfaces) - i can not. There is no clone options - i have to add it manually typing all 100 Items with all parameters like timers, units, history...

You're right it's not called "clone" it's called "copy selected to" and it's in the dropdown at the bottom of the list of items and works on the items you have selected to "copy to" another host or template.

h) cisco templates from: http://www.zabbix.com/wiki/templates/start - none of them works (no error durring import, no errors in logs but the template does not appear after import). Existing templates for cisco: 87x, 2960, old pix 515/525 - it's definitely not enterprise gear.

Perhaps you can be a little more descriptive of you issue here. Many of those templates are submitted by people in the community, I'm sure they worked great for the people who took the effort to create them for their own uses and then felt a desire to share.

Of course i found some other bugs, but it's not the issue here.
The summary for zabbix is: it's not for enterprise, and it's good only for monitoring PC stations using zabbix agent - definitely not for network gear, and other SNMP monitored gear. Moreover zabbix archtecture is not flexible and is much worse then other opensource available on the market.

I think it was Alan Cox who said "All software sucks." I'm sure someone somewhere said something like "All software has bugs."

So now that we know that all software has bugs and sucks, I am very thankful that Zabbix SIA is very bad at making vacuum cleaners because Zabbix definitely sucks a lot less than other software out there. <caugh>Nagios<caugh>

I really appreciate your work, i know it's very difficult to spare your free time to build such opensource. But let's face a truth. Am i wrong ?

Hmmm... are you wrong?

I'll just leave it with, I disagree with your conclusions.

I hope some of this has helped understand Zabbix a little better and I look forward to you using Zabbix in the future.

[highland]

Hmmm

I gues i was wrong
Your tips solve some of my issues, it seems much better now!
But still i have some problems:

I have one template for my 10 switches, the problem is in Item description which is being drawn as a legend on a graph. The switches which are using this template are identical with ports GigabitEthernet0/1 to 48. The problem is that those port numbers on graph do not tell me anything. I need to see port number along with it's description (OID: ifAlias, for exampe GigabitEthernet0/5 - database5). I've tried to put in Item description: GigabitEthernet2/1 ifAlias["index","ifDescr","GigabitEthernet2/1"]
but its not being processed (on a graph it's drawn as a text).
Do you see any other solution for that ? Maybe it would be good if i could override Item description in Host ? (right now it's "greyed" when i try to edit Item description from Host).

Sorry for previous frustration, it seems that i have to do much more reading (and asking on this forum)

Best Regards,
Michal

[angelhc]

Hmmm
Sorry for previous frustration, it seems that i have to do much more reading (and asking on this forum)

Hello Highland,
I think that Zabbix is the better monitoring open source application after one year evaluating all of the monitorization tools on my work team (not me alone).

Actually we've 11671 items monitored on 250 servers and networks elements (yes, not too much but we implement the application on Aug. 2010) we're generating 128 checks per second which trigger 6469 actions.

Really you think that's not for enterprise management systems?, that's only 5 months of work and look at our growing environment, all of them are from production ones.

All of your dissapointments are caused by inexperience managing the tool, and of course there's a lot of bugs, must to do and whatever but if "all sucks, all is wrong, all have bugs" Why not you improve it and share to the community

With best regards.
Angel.

[highland]

OK, right now 2 problems left:

1. Description of Items (from Template) should be "dynamic" (described in prev post), or at least changable at Host level. Any suggestions for that ?
2. Is there any way to make zabbix syslog server and to create triggers based on received syslog messages ?

Thanx

[Colttt]

1) {Hostname} =its a variable.. and would be change by zabbix automaticly to hostname..
2) add an element: Zabbix-agent (active) and you will find eventlog, log and logrt i think this is what you mean

[ufocek]

2. Maybe you can try this software:

http://www.zabbix.com/forum/showthread.php?t=19180

[nelsonab]

Threads like this make me smile. :-)

Highland I'm glad to see that the community has been able to take some steam out of your frustration, hopefully soon we can replace it with excitement for Zabbix.

Your issue of the graph titles not having descriptive names is the edge case I was referring to. I made a patch for it back in the 1.4 days but it required some DB changes and unfortunately it didn't get included into the source. This however is an edge case that when you hit it is very annoying.

The main work around for it is via screens. Either above or below the graph you have a cell which contains text with the human readable description.

Another option might be to create your own graphs where the title may reference either a user defined macro or some macro value inside one of the items.

wooo hoo post 999!

[highland]

@ufocek - thanks - i'll try this solution tomorrow. Does anybody know if this functionality might be integrated in new version ?

@Shad0w - can't install agent on cisco routers

@nelsonab - i have routers with 20+ vlan interfaces and i would like to see all the intrface traffic on one page without addidional mess - i suppose your patch does not work for version 1.8 ? Tomorrow i'll try your solution with screen&text description and check if it looks ok.
EDIT: @nelsonab - i've checked it - it looks fine !

More:
1. i would like to configure zabbix as snmp trap receiver - and have specific actions for specific triggers (which are specific snmp traps). Do you know how to do it easily without to much configuration on snmptt side ?
(or rather such configuration which gives me flexibility and possibility for network operators to add new triggers based on traps easily)
Another functionality i would gladly see integrated into zabbix.
2. I have problem with external script:

zabbix@sparc /root $ /etc/zabbix/externalscripts/netout.sh
2670441195
zabbix@sparc /root $ /etc/zabbix/externalscripts/netin.sh
367251059
zabbix@sparc /root $ id
uid=1005(zabbix) gid=1008(zabbix) groups=1008(zabbix)

I defined item with type=external check, and key=netin.sh (or netout.sh), tore-value=delta. The problem is that it sometimes it gets good values, but after a while it's not supported. Logs says:

Item [router1:netout.sh] error: Script /etc/zabbix/externalscripts/netout.sh returned nothing
Parameter [router1:netout.sh] is not supported, old status [0]
Parameter [22919][router1:netout.sh] became supported

This process repeats, once it's supported, few minutes later not. Graph is sometimes drawn (propably when it's supported) - but incorrectly.
Both Items has multipler 1000000 (large numbers). Where is the problem ?
EDIT: Solved: problems with permissions in some log files

3. Problem with maps: there is no icon for hub and router - they look like network icon.

4. Another issue with maps: can i define my own icons ? I would like to have one icon when device is OK (for example with green background) and other icon when some trigger fails for that device (with red background). Icon could be changed on trigger. Is it possible ?
BTW: it would be great if also the links where tied to Items (for example SNMP interface status) - this way i could also knew which link has failed (routers/servers has many interfaces)

Thanx

[Colttt]

hehe.. you right.. sorry

i´ve HP ProCurve-switches, via SNMP i´ve become the name from the switch it names sysname..

perhaps you can tell zabbix that the sysname = {Hostname}?!

[zabbix_zen]

A little short on time,
but for your 3) and 4) questions,

you are able to upload your own icons navigating via
Administration | General
and selecting Images from the upper right corner drop down list.

As far as I remember it's possible to change the link colors,
check this