Call for discussion on ZABBIX 1.8

[troffasky]

- When an item is defined, it can be stored "As is", "Delta (speed per second)" or "Delta (simple change)". It would be interesting to create one item stored "As is" and define two "computed" items as item_speed=delta_per_second(item) and item_variation=delta_simple(item).

Yes, this would be useful, especially [for example] as data on interface throughput is gathered in octets [bytes] as a continually-incrementing counter but is much more useful to the end user as bits. Even being able to create an item on a graph that's calculated [so ifInOctets1 * 8 to get bps rather than Bps] would be useful. You could even add a few macros for commonly-used functions [bytes to bits, for example].

[troffasky]

Pick objects from MIBs when creating new items on a host/template

When creating a new item on a host or template, it would be useful to be able to browse the MIBs installed on the system, and select objects from them to add to the host. The DESCRIPTION field in the MIB would be useful to see as well. Being able to hover over an item and see a tooltip with the DESCRIPTION in it would be good. For example, in the ifMib, the DESCRIPTION of the ifInOctets object is:

"The total number of octets received on the interface,
including framing characters.

Discontinuities in the value of this counter can occur at
re-initialization of the management system, and at other
times as indicated by the value of
ifCounterDiscontinuityTime."

If you could upload MIBs through the web interface, that would be even better.

[stephen.soltesz]

I wanted to add one more vote for an xmlrpc API and a clarification of how I would use it.

I can see how a comprehensive API would be a massive task, but my guess is that it does not need to be comprehensive. For instance, initially it could be an administrator-only interface.

I would like to extend the import/export functionality for configuration information other than that currently supported by the GUI. Actions, users, and user groups, and what hosts they have permission to modify come to mind immediately.

This would allow me to quickly configure a new zabbix installation using an external host and user database, and start off using predefined actions for known scenarios.

Thank you!

[marcis]

+1 for more flexible items

This is a hard one to get right. Anyone who's used a Mitel 3300 web interface will know what I'm talking about. What I'd like to be able to do is add multiple objects at once, and increment the values. For example, I'd like to add ifHCInOctets1 - ifHCInOctets48 to a template. This would involve incrementing at least three objects per item added [Description, OID, Key], which makes it something of a tedious data-entry job. Obviously it doesn't have to be limited to incrementing - I'd want to be adding In and Out values for all the given keys, so being able to do that at the same time would be useful. In bash, I can express it fairly simply:

for ((i=1; i <=48; i++))
     do echo ifHC{In,Out}Octets$i
done

When executed this will expand to print all the strings I want. So, how do we translate this to a web interface? Right now, if I want to do things like this, I

1. - create the first objects of the sequence
2. - export it to XML
3. - pull out the fragment of code I want [eg the <graph></graph> bit]
4. - slap it in a bash script [ie put it after the 'echo' in the script above]
5. - substitute the bits I want to be incremented with $i
6. - run it, redirecting output to a new .xml file
7. - paste the rest of the <zabbix_export ...> code around it
8. - re-import it into Zabbix

Phew! Tedious, but nowhere near as tedious as manually creating every single one. The 'Clone' button is a step in the right direction, but unfortunately it's only one step, and some operations like this could take thousands of steps [48 ports on a switch, 12 objects per port, you get the picture]. Lets just call it an anti-RSI tool :-)

I also find it very unpleasant to create 48 templates for each port of some switch. Maybe some mapping thing should happen ... where associating host with template would additionally require user to enter the port names which is mapped to appropriate OID's. Mapping would be necessary because there is devices which have ports of various type (e1, e2 for Ethernet, g0 and g1 for GigabitEthernet or even VLAN) which resolves to different ranges of OID's (e1-e48 uses 1-48, g1-g4 uses 49-52, vlans use for example 2000-6095)

[cjwallace]

I would really like to be able to use Zabbix from my Blackberry or Windows mobile.

I dont need to have all the functions working but it would be nice to see whats up whats down and maybe run a ping \ trace route

Please Please give mobile devices some thought as alot of other products paid and free have this.

Craig

[noxis]

Creating graphs from Triggers as well as Items.

What I mean by this is you create a set of triggers on a network interface:

nic1 > 5mb
nic1 > 20mb
nic1 > 50mb
nic1 > 75mb

Then you can create a pie chart based on how much time that trigger spent in any of the 4 states. I guess this is to create a wide range of SLA based graphs. You define your SLA's as a triggers and then graph on them for reporting purposes.

[troffasky]

SNMP auto discovery

I also find it very unpleasant to create 48 templates for each port of some switch. Maybe some mapping thing should happen ... where associating host with template would additionally require user to enter the port names which is mapped to appropriate OID's.

It should not be necessary for the user to enter the name of an interface - the name of the interface is available in the ifMib as ifDescr. When creating graphs for example, I'd like to be able to reference the value of an Item in the Name of the graph. So instead of creating a graph called "Interface 1 Traffic", I'd like call it "{HOSTNAME}:ifDescr.1 Traffic". That way, if device in question reports a sensible value for ifDescr.1 like "WAN" or "Dialer0" or "Link to Cabinet 15", then that's what appears in the graph. The same principle could be applied to Triggers, and probably other areas too.

Mapping would be necessary because there is devices which have ports of various type (e1, e2 for Ethernet, g0 and g1 for GigabitEthernet or even VLAN) which resolves to different ranges of OID's (e1-e48 uses 1-48, g1-g4 uses 49-52, vlans use for example 2000-6095)

Yeah, tell me about it! Different vendors use different indices for their interfaces. For example, interface 1 on HP Procurve is .1, on 3Com is .101 and on Cisco switches is dependent on speed, so Gi0/1 would be .10101. What's needed is some kind of auto-detection when adding devices that don't have templates. Templates are all very well on homogenous networks, but where you've got a mixture of vendors and models to monitor, things can get tricky.

Template auto-creation = more devices supported = more users.

[troffasky]

Please Please give mobile devices some thought

What part of it doesn't work for you? I'm using 1.4.6 fine with my Nokia E61's bundled Webkit-based browser. The graphs are even sized to a sensible width. Perhaps you need to use a different browser? When I say 'use', I'm just using it to view the triggers/latest data/graphs.

[BHG_2008]

Greetings!
Here's my wishlist.
1) FULL import/export functionality, to the extent the export could be used as an emergency recover.
2) Add LDAP support for Groups (create group in ZABBIX and AD; manage group membership from AD, permissions of group from ZABBIX?) (Make Domain Admins happy)
3) Replicate users and user groups and their permissions in DM configuration.
4) Fix broken permission code which does not merge permissions from each user group, as this does not allow combinations of membership to have effect. (Reads and applies first user group encountered, ignores subsequent matches).
5) Add Filter By Host and Date Range to Events
6) Add Filter By Recipients and Date Range to Actions
7) Document new configuration file parameters supported, sorted by release.
8) Add failover Master server parameter to zabbix_server.conf, for use in DM setups.
9) Separate History and Trend expire timers from Items, to give DM Masters INDEPENDENT control of how long each are kept. In my scenario, I need a Master the NOC looks at for alarming by application (requires consolidated data to determine), and another Master for my use as Capacity Engineer to do trend data analysis. First would need no trends, and very little history, second would need no history and trends for 13 months!

Much thanks for 1.6! Can't wait for the DBSyncers in 1.6.1 to see performance improvements of 4-8 times as advertised.

[BHG_2008]

Discovery Modules

Need to add one more, Alexi said BIG ideas, so here goes:
Add plugin module capability for the Discover feature.
Add a database table which houses Discovery defaults by module.
Then, the userbase could contribute modules just as they do templates today.
The purpose of the modules would be to use external code, preferably written in php or perl, which would employ SNMP to query hosts found by discover, and automatically set up the host entries for the devices, conditionally link to templates, add custom or template-independent items found. My associate has already written one such widget, which queries switches, routers and just about anything with a network interface, adds the host, sets up interfaces found as items for up/down, errors in/out, and descriptions, and threshold triggers as you specify in the perl file. This generates an XML file, which brings us full circle back to the module. Instead of outputting to a flat file, the module would simply call the import routine in the Import/Export feature. Done.

This would result in modules which could be built and certified for use by ZABBIX, by release, so that development/expansion could be added between ZABBIX releases.

Just publish guidlines the modules must comply to, such as looking up defaults from the Discover table, installation of module must include default table population, as well as follow the XML tag formats already in use by ZABBIX.

Big enough?

[gospodin.horoshiy]

I think this is the right place to post this:
Here is what I'm missing most at the moment:
1)Full UTF-8 Support
2)Daily Reports(also via mail) - since many alerts are not so critical and can wait to be examined all at once in the morning.
3)Inventory Autofilling.
4)Ability to send 'remote commands' to hosts who do active checks only.
5)Windows Proxy

Thanks for the great product Alexei, Aly and co!

[richlv]

...automatically set up the host entries for the devices, conditionally link to templates, add custom or template-independent items found.

actually, this sort of thing seems to be something that would sit well on top of an api, if we ever see one
otherwise it seems to be pretty hackish and doomed for reimplementation on top of the api anyway.

[Fatima R R Tonon]

IT SERVICES for an specific period

I was wondering if there is some way to make the IT Services show not only the last week SLA, but let us enter the initial date and the final date.
This would make easier to everyone calculate the SLA for any day of the month or for an specific period...

[alixen]

Re: IT SERVICES for an specific period

I was wondering if there is some way to make the IT Services show not only the last week SLA, but let us enter the initial date and the final date.
This would make easier to everyone calculate the SLA for any day of the month or for an specific period...

I have modified IT services code to display SLA for last N days (instead of hardcoded 7 days) and it works. It is currently an "ugly hack" that needs some polishing before being published but yes, it seems quite easy to produce an SLA for a specific period of time.

[data7]

My suggestions

Well I've been using Zabbix for quite some time, loved the 1.6 release (although it still does have a lot to improve) so here goes my feedback:

Let's start with the pros, Zabbix's administration offers a wide range of possibilites to organize stuff that is even hard to see on non-free solutions and besides that it is also very efficient on collecting data (I already got unavailability notifications that took 30s for example).

- As great suggested, I also vote for an API. It would be better for adapting the solution to the real necessities of a company.

- Integration to other well-known tools - Zabbix is a new player on the game and in most situations we will be facing a migration. On large-consolidated environments convincing a company to change is quite hard. It would be easier if integration with Nagios existed (probably not Nagios itself, but AT LEAST the plugins like Zenoss does), Cacti (for the SNMP area), OpenAudIT (Inventory) and one for the report like someone said before me. It would be VERY nice too if it could take some advantage of NMON output data for Linux/AIX hosts, since default performance collecting (especially for AIX) is very poor.

- Zabbix comes with lots of checks by default but it should have some automation. I evaluated Quest's Foglight (http://www.quest.com/application-and...ce-management/) which is an expensive but VERY powerful monitoring tool built on Java/Tomcat that uses modules (called "cartridges") for enabling its funcionalities (and obviously you pay for each one). It does incomparable monitoring of host resources (on most operating systems by auto discovering procs, filesystems, and a lot of other performance data), applications/DBs (VMWare, WebSphere, very easy setup and complete analisys for Oracle,DB2 etc) and Web monitoring with Rec'n play features for end user experience. If Zabbix could get some automation (by querying the host) to get its information it would surely be one degree ahead from all monitoring tools.

Now for the little adjustments:

- RPM/DEB compiled packages like Zenoss that include all library needed by the software. I had some trouble dowloading libs through yum before, this may have come handy.

- SNMP Wizard, like Cacti/Zenoss does

- Maintenance mode for hosts/hostgroups and downtime scheduling (no triggers would popup during such times). This shoudn't be so hard so if you could release it on 1.6.x series i would be VERY glad.

- Actions for "Unknown" states. I run into a daily routine of checking if all agents are working properly. It would be nice to be notified when an agent stops working for x period.

- Database Wizard. It isn't so hard to setup the database for Zabbix but some users experience trouble on this. I believe more users would use Zabbix if it were a bit more easy to install.

- More filters. It's one of Zabbix greatest features but it still have to improve. Remember Zabbix developers, there never enough filters on a page! Some time you are asked to see stuff time related and filters don't help. The "Item" section had improved greatly from 1.4 to 1.6.

- I also would like to see how Zabbix does things. Ok, it has a lot of pre-defined parameters but sometimes I experience trouble with them. It would be easier to debug the error if I knew EXACTLY what command-line is used for checking.

Well, that's it for now, if something I said here already exists on Zabbix please let me know outside the topic (e-mail, PM).

Thanks a lot and keep up the good work!