Call for discussion on ZABBIX 1.6

[StanZoid]

Another thing I would like to see addressed is log monitoring. As I see it today, the logs are loaded into the database, then the triggers are applied. In my case, where I am monitoring tens of logs on a single server, this creates a lot of unnecessary database action.

This leads me to two types of log monitoring. The first type is only interested in finding defined strings and alerting if they are present. I see this as an agent-level activity, with only results returning to the database. It would be much easier than generating a script and using sender, or defining a custom variable and hoping the delay is under 1/2 second...

The second type is interested in the contents of the log, such as a cron history. The current log functionality works in this case.

One other thought on log monitoring. Many of my logs use naming that includes date stamping, making log checking impossible to configure as currently implemented. I would like to see an option for log monitoring that considered variables in file names.


Great work on 1.4!
StanZoid

[bking]

Two things I noticed in Zabbix that are incomplete that make operations cumbersome:

1. The backup system doesn't backup everything necessary to have a working system on restore. I'd like to press backup and have everything except possibly logs. When I take this backup to another machine (or a freshly reinstalled machine) I'd like to be able to press restore and have my zabbix running again. The last version I tried for this was 1.3.8.

2. Editing creating a map is cumbersome when setting it up because the coords have to be typed in to move an item. Here it would be much more user friendly to enable drag and drop with the mouse to adjust locations on the map.


Another user friendly option would be to add a button perhaps to the acknowledgments area to increase or decrease checking time for individual triggers. This would allow to quickly fine tune if triggers are happening to fast.

[Andreas Bollhalder]

Auto {HOSTNAME}
My great wish is that ZABBIX is automaticaly adding the hostnames to the triggers, maps and other location where needed, removing the need to prefix every trigger with it.

Selection for media type in actions
It would be nice to define different actions depending on the media type of the user. With this, it would be easy to setup the message text different for example for sms and email.

Andreas

[ghislain]

Here is a few things:

Device discovery:

- disk discovery : We want to monitor disk inode, and free space on all server. Right now you must know each server's mount table to monitor it and create triggers, item and action to suits the partition schema of each server.

It should be easy to zabbix to have rule like :

vfs.fs.inode[/,pfree,all]

that does the monitoring on all disk. Same for all disk related functions.

- ethernet card discovery : same as disk discovery but for ethernet cards.


LDAP user authentification:

Enterprise size have allready a user directory so plugins like LDAP auth or Mysql auth should really make the thing easier.


API :

API is also very usefull when you have an enterprise system, how could you manage all those servers if they are in 5 different applications. The monitoring should be easy to plug into the schema and so provide API for adding/removing servers at least equal to autodiscovery possibilities.

see zenoss for all those features. Seeing them in zabbix would be great.



Automatic type for items:

Nothing frustrate me more than realizing a zabbix item was not working because i let an item type default to integer when it should be 'float' or other.

This should not be possible for zabbix agent and for internal check. Those type should be hardcoded in the gui to prevent any errors and frustrations




Distributed monitoring:

One thing to manage is false alarm. A great way to have this remove is to have checks like this:

do the check from server 1, if all is ok then proceed
if this triggers an alert then do the same from server2, if this also trigger then validate the trigger.

It means the alerts has to be confirmed by a second server



thanks for zabbix. Hope that these sugestions will help


regards,
Ghislain.

[bibi23]

My suggestions for next version of zabbix:

Concerning DB architecture:
* first, i think data integrity with foreign key is very important.
* Second, all database don't have the same level of support (oracle not supported in 1.4, not transaction support for postgresql, watchdog only for mysql....). perhaps with odbc, it should be more easy to have the same support and to not spend time with different database driver.

Concerning fonctionnality:
* it should be fine to have a test bouton when creating a new media type
* negative graph
* move snmp info to an host level or enterprise level (not in item)
* the possibility to add a trigger in graph (not only in latest data screen)
* in screen, the ability to add trigger overwiev for only one host
* in screen, if there is a lot of graph, it is difficult to find the right graph in the select menu. splitting this select into multiple select ??
* AJAX / WEB 2.0 ????
* and the most important thing, a wizard or equivalent to create new item:
when i create item, only item avalaible for the host OS should appear in select; and for a selected item, only avalaible counter should appear (for example: sda and sdb if vfs.dev.read is selected) and the corresponding unit should be set automatically (integer, float, log....)

Concerning doc:
* IT Services is not clear: an example should be great.
* Log monitoring for windows is missing
* the winpopup chapter should include code and not the hyprelink to forum.

else, the zenoss web site is interesting. http://www.zenoss.org/

Thanks

[bking]

I just found out one that very much needs to be added.


Protection from removing the admin superuser account.


Yes, every once in a while it is nice to have a backup.

[LEM]

1.6 improvements - suggestions

I would suggest spending some time to stabilize ZABBIX code: using automatic testing discipline to have the whole zabbix code (php frontend and daemon/agent code) under automatic tests. This would helps you in the future. In fact, Zabbix as achieved a lot of new features since 2K5, but suffers from bugs that could be detected by a stronger testing discipline.

I would suggest too implementing an API for smoother integration with third parties softwares. The kind of API a other application, php or not perform actions that can be done by the user interface (crude-ing hosts,groups,users, and later enabling/disabling items/triggers/actions, crude-ing maps...).

Working on SLA based on "real life" SLA schemes could be something not too bad too. For that, perhaps calling for participation for SLA sample submiting could helps.

[monsieurcanard]

Could the 'stable' version be updated to say 1.3 or 1.4, and then maintained as a bug-fixes-only branch?

Also, I've been thinking of making a patch to plot servers (and their status) in a Google maps mash-up. Would anyone be interested?

Most of all I'd like to congratulate Alexei for such a brilliant product. Zabbix really has come on leaps and bounds since v1.1.

Matt

[monsieurcanard]

PS I'd like to second the built-in SSL/authentication idea. I'm currently running Zabbix agents in active mode and have them report back to the server over OpenVPN, but it's not ideal (I can't use it for client dedicated servers) and I currently wouldn't dream of allowing remote commands in Zabbix agent because of the lack of built-in SSL/authentication.

[StanZoid]

Action configuration

Since I now have my monitoring completely defined in templates, I now need to configure actions that (A) alert to specified media (B) based on the template containing the trigger that fired.

Stanzoid

[LEM]

PS I'd like to second the built-in SSL/authentication idea. I'm currently running Zabbix agents in active mode and have them report back to the server over OpenVPN, but it's not ideal (I can't use it for client dedicated servers) and I currently wouldn't dream of allowing remote commands in Zabbix agent because of the lack of built-in SSL/authentication.

I also think it's should be a good thing to have strong authentication for zabbix agents/servers scince active mode is probably widely used (and damn, really usefull .

Sure SSL is somewhat visible... so could be a 'marketing compliant' solution too.

[Alexei]

Thanks for all suggestions! Please keep going. I will come up with my comments just after release of 1.4.1.

[dratliff]

Could there be a way using the trapper to also process a push of a timestamp with the data? Or does that already exist and I am ignorant of it?

If you could push historical data into zabbix based on a timestamp, as well as data, that would greatly improve some of my monitoring needs. Specifically, asynchronous database statistics based on start-end date criteria from multiple database platforms or remote units.

[Alexei]

If you could push historical data into zabbix based on a timestamp, as well as data, that would greatly improve some of my monitoring needs. Specifically, asynchronous database statistics based on start-end date criteria from multiple database platforms or remote units.

I assume that the historical data should not trigger any events, it is just for logging, right?

[dratliff]

I assume that the historical data should not trigger any events, it is just for logging, right?

Initially, the data is just to display in a graph on a screen. Sort of broad troubleshooting scenario. I am going to assume that after we get smarter on what we really want we are going to want to warn someone that the response time from a unit is unacceptable or something to that effect.

I am going to query a vendor database every so often and get new values. From that data I am going to extrapolate the communication rates (initially). I want to timestamp an initial unit query request, server query response to the unit, the time differential and then the unit data request, response, etc. until the end of the transaction to the unit. It is mainly a quality assurance and performance analysis.

I'd like to display this data with server (local and remote) data collection on a screen so that we can determine if there is a problem as well as log a variety of conditions for troubleshooting.