Call for discussion on ZABBIX 1.8

[richlv]

Perhaps this thread should be turned into a wiki? Given that everybody's got something to say about everybody's ideas, a wiki would be a good format for discussing new ideas; the linear nature of a forum thread makes it difficult to follow different 'branches' of ideas.

well, http://www.zabbix.com/wiki/doku.php?id=contrib:ideas
it's very hard to manage a list of this size in dokuwiki, though - i'm used to mediawiki, and i miss quite some features. supposedly some of them can be implemented using plugins, so i'm waiting for that to happen.

[arli]

Sso

Support for SSO when user is already authenticated using apache-kerberos, then REMOTE_USER variable value could be enough to grant access to according Zabbix user.
I belive it should be minor change from HTTP authentication, in which case password is also expected.

[drvo]

Suggestion on V1.8 features

Hi,

I worked with IBM Tivoli and HPOV monitoring products for 10 years. I am using Zabbix 1.6.2 for 2 weeks and it is great but I am still missing some functionality. Here are my suggestions:

1. documentation, documentation and documentation with examples especialy for discovery, zabbix_sender (-h produces non clear instructions)
2. agents on different platforms should have at least the same funcionality out of the box (monitoring CPU, RAM, HDD items) and of course out of the box templates with same type of monitored items, triggers and graphs
3. Monitoring filesystems on linux machines should be based on "autodiscovery" of all mounted filesystems - it could be implemented in this way: thru inventory scan when the node is added you can discover CPU, disks, ..., then on Configuration > Items you could have a link similar to [Show disabled items] ie. [Show discovered items] and there would be prepared this items for the host
3. Monitoring filesystems on linux - if something is unmounted, it should report "unsupported" or something like that and not report fake numbers from underlying filesystem !
4. Discovery should add hosts based on ping availability - baybe this functionality is already there, but then it should be better documented (i could not get to install fping on my server, maybe this is the reason, could you insert the fping code to zabbix server ?)
5. Trigger creation in one step with item (it would be nice to have the posibility to create 2 treshold values i.e. 2 triger creation in one step when the item is added/edited - warning/critical treshold)
6. If event is created thru trigger - event description should show the value of the monitored item
7. I agree with alixen with "Computed items" posted on 23-10-2008, 14:14
8. Support for monitoring of virtual environments (VMware ESX thru SNMP template, XEN, ...), ESX and XEN guests machines could be added thru inventory of the host server.
9. Last but not least, please add support for remote WMI monitoring of windows machines at least in one of this ways:
- zabbix_wmi_proxy that could be run on windows machine and collect the preformance data from other windows machines
- zabbix_wmi_sender or enhance zabbix_sender that could periodicaly check folder with csv data from performance monitoring on windows "Performance Counter Logs". On windows you could set remote monitoring of bunch of servers and log the performance data to csv files. This data are needed "only" to be imported to Zabbix with the time they were logged (i.e. zabbix_sender should support import of time;key;value "pairs")

Thanks for Zabbix, well done.

[richlv]

I am running 1.6.1. I know the only thing I changed in this case was the item title and clicked save.

hmm, it seems you are right. the functionality is available in trunk, but not in 1.6 branch - hopefully it will be merged/backported soon

[verwilst]

The thing i would like to see more in 1.8 is trending. Being able to anticipate to problems is gold

[Andreas Bollhalder]

I would like to see the SNMP credentials are configured by host like IPMI and the integration of the SMASH protocol.

Andreas

[alkeksibus]

we would like to have some more visualisation features within maps, i.e. showing the host ip address next to an icon as well as it´s label.

[sundaysteve]

zabbix wish list

-smarter UI, making things look nicer will go along way to getting more people interested in zabbix

-ability to intergrate in ticket loging systems like vTiger CRM, when an alert is trigger under certain conditions a ticket is generated in another system, most CRM system can generate tickets from a simple email provided you can formate the email appropriatly.

-simpler agent install/setup process & remode agent administration

-clearer roadmap/timeframe with more frequent updates

Greate work on zabbix so far!

[bcheese]

Need to reduce network noise.

Hi Alexei and Co,

We have a need to reduce the number of network connections being performed to monitor our hosts. The problem is that I have say 8 servers in a DMZ environment which have about 50 monitored parameters each. If (and this is not quite correct) each of these parameters are requested every 30 seconds, then I am looking at 1.152 million TCP connections through our firewalls per day.

Now because these hosts are in a DMZ we are required to log all connections in and out of the servers in question. As you can appreciate seeing 1.3 million connections per day where most of them are the health monitoring is a little bit of a problem to deal with.

To solve this problem I would propose that the zabbix_server and zabbix_agent{,d} is altered to include a few smarts to allow the server to operate in the following maner:

1. Obtain host to check parameters on
2. Connect and authenticate with agent
3. Request first parameters value
4. Check to see if there are more parameters to collect. If so, then go back to 3
5. Disconnect from agent
6. Resume current normal operations

The proxy in its current state provides no help here as the DMZ hosts are not allowed to connect back into our corporate network.
This change in itself would make my network administrators very very happy.

[Alexei]

Hi Alexei and Co,

We have a need to reduce the number of network connections being performed to monitor our hosts. The problem is that I have say 8 servers in a DMZ environment which have about 50 monitored parameters each. If (and this is not quite correct) each of these parameters are requested every 30 seconds, then I am looking at 1.152 million TCP connections through our firewalls per day.

Now because these hosts are in a DMZ we are required to log all connections in and out of the servers in question. As you can appreciate seeing 1.3 million connections per day where most of them are the health monitoring is a little bit of a problem to deal with.

To solve this problem I would propose that the zabbix_server and zabbix_agent{,d} is altered to include a few smarts to allow the server to operate in the following maner:

1. Obtain host to check parameters on
2. Connect and authenticate with agent
3. Request first parameters value
4. Check to see if there are more parameters to collect. If so, then go back to 3
5. Disconnect from agent
6. Resume current normal operations

The proxy in its current state provides no help here as the DMZ hosts are not allowed to connect back into our corporate network.
This change in itself would make my network administrators very very happy.

You may consider using active checks along with custom settings for agent-side data buffering. This will significantly reduce number of connections. Proxy is another alternative...

[bcheese]

You may consider using active checks along with custom settings for agent-side data buffering. This will significantly reduce number of connections. Proxy is another alternative...

Thanks for the suggestions Alexei, however the devil is in the detail. I cannot use active checks for these items as they are in a DMZ networks. Since when have the agents been able to buffer parameters? I have seen nothing in the manual about it.

Proxy is not an alternative for me as again we are talking about a DMZ (unclassified) network being monitored from a classified one. The proxy would become more useful for me when it allows the connection in reverse as per a previous suggestion for 1.8 I made.

The proxy in a DMZ supported by data buffering active checks may help, however I still have the problem of an unclassified connection into a classified network. I will go and re-read the manual and see.

Cheers,
Brian.

[tmine]

About Actions for Discovery

Hi, all.

I think it will be help to have "Disable host" as Operation Type sellection @ Action Opereation.
Because it will be useful if I do not want to "Remove host" but just would like to stop monitoring for missing host

Best regards.

[richlv]

Proxy is not an alternative for me as again we are talking about a DMZ (unclassified) network being monitored from a classified one. The proxy would become more useful for me when it allows the connection in reverse as per a previous suggestion for 1.8 I made.

i might have misunderstood something, but what's wrong with proxies in your situation ?
you place a proxy in dmz, which gathers information. zabbix server connects to the proxy and gets the information.
a quick search turned up a somewhat old, but more detailed post

http://www.zabbix.com/forum/showthread.php?p=32846#post32846

[bcheese]

i might have misunderstood something, but what's wrong with proxies in your situation ?
you place a proxy in dmz, which gathers information. zabbix server connects to the proxy and gets the information.
a quick search turned up a somewhat old, but more detailed post
http://www.zabbix.com/forum/showthre...2846#post32846

Hi richlv,

What you described is how I want it to work, but it is not how it works. My problem is Alexei's point 2. In short I need the connection to only be Server to Proxy.

2. It requires one-way (Proxy to Server) TCP connections only.

The problem with the proxy for me is the security manuals under which I must operate (Australian Government) do not permit network connections from a lower classified network (my DMZ's) to one in a higher classified network (location of my Zabbix server). As a result of this the proxy in it's current form is pretty much useless because it relies on being able to make a connection from my DMZ's back into my zabbix server (TCP port 10051). If I were to allow it then I would be in breach of security protocols and risk loosing my job.

So, as a result of discounting the proxy which is a great concept and I have already made the suggestion of allowing the active connection to be created in the opposite direction, I then looked further into the manual which is now a lot better that it has been in the past (read well done Alexei) with a view to trying to use the active protocol for reducing network connections, but guess what. The active protocol has the same connection direction problems which the proxy has. So that rules out that option.

Cheers,
Brian

[linuxdan]

What about 1.6

I really think more time should be spent on making 1.6 work right before diving into 1.8 I'm having issues with the following areas of 1.6:

* User parameters
* Macros in notifications
* Server Inventory
* Host Availability
* Inadequate debugging

etc....

I'm running 1.6.2 on RHEL4 and RHEL5 servers.