I completely agree! About 80% of the graphs we use are like this.
I create a group 'Webcluster' with a bunch of templated hosts in them, all with simple graphs.
Now all I want is for example the Simple Graph CPU load for all 'Webcluster' hosts in one graph.
This costs a lot of RSI and time to build & maintain, while it would be so logical to add this in anything but the smallest environments.
This is one of the very few features I'm truly direly missing in Zabbix (I now have stupid-ass shell scripts multiplying XML's to do it for me), and it would save a lot of people a heck of a lot of time (and mouses) to implement it.
I very strongly support this RFE.
Back link from originating thread: http://www.zabbix.com/forum/showthread.php?t=20093
This gets my vote! We would strongly benefit for this feature as I am trying to set up a "monitoring wall" so our general users can also keep track of several things in our infrastructure - this kind of feature would make my life so much more easier! 
Any expectations on this highly wanted feature?
it's not on the 2.0 roadmap, which can be considered to be a relatively short term roadmap
Need this all the time. Currently manage these manually using calculated check on a fake host pointing to actual hosts and create a graph from that.
Just adding my support for this - it would be very useful for us as well.
X1000. My boss won't allow us to fully transfer until this feature is implemented. For the love of all that is holy, help me get off of nagios and implement this feature.
note the part about "me too" comments and voting at https://zabbix.org/wiki/Docs/bug_reporting_guidelines 
lots of almost identical comments make it hard to work on the issue later.
also, if somebody knows of a company that wants this feature enough to finance it, just use http://www.zabbix.com/contact.php
ZBXNEXT-1120 is about displaying a specific graph for all hosts in a group
What is some of the leg work the community can do here to make this a feature a higher priority.
Management would love to be able to see the entire group of web servers behind the load balancer and see the graphs on them, right now... that is not easy to do.
convincing the management to finance a specific feature tends to work ok
In the hope that it helps someone else, I have published a Perl module I wrote to help in creating graphs containing item X for all hosts in a host-group.
(1) this would be even more needed for lld graph prototypes - maybe as a graph to include all items with identical keys, all discovered items, or maybe all item keys matching a regexp.
(2) should work with host prototypes (vmware discovered hosts for now)
Any help on how to use this script?
We need that feature.
Hey, we are in the cloud era so it's really hard to track host created by users and do graphs for them. And after a while some just get power down and new ones power up and we start from beginning. So we need a group graph feature so we can graph all linux group servers memory for example. That goes with host discovery.
Specification available at https://www.zabbix.org/wiki/Docs/specs/ZBXNEXT-570.
Can we have some sketches for new page? I think description of new feature is good, but sketch would help a lot to visualize this.
Hello Pavels, thanks for your update, that would be a very welcome feature!!
I am however a bit puzzled by your description:
> Graph page
> The graph page will display the graph with the selected items.
> If only one item is selected, the page will be displayed as a normal simple graph page. Otherwise, the following changes will be made:
> to avoid displaying too many lines on the graph, only the average item value will be displayed.
(...)
If i understand your description correctly, this would effectively mean that you don't see the data from all the hosts you selected in the filter, but just an average... which is almost useless to quickly ascertain the health state of a group of hosts at a glance!
What i am really missing in Zabbix (and i think many bug reporters here also) is the possibility to view all the hosts of a group at once, as in Ganglia, say:
http://www.dotsch.de/boinc/DotschUX-Ganglia.html
So all CPU-loads in the group, ideally in separate small graphs. All values together sharing the y-axis in the same graph could be perhaps offered as an option, but as you noted it becomes messy for more than a handfull of hosts (say, more than 5-10 hosts).
Actually, that is wish-request ZBXNEXT-1120 .
Would it be possible to address the two "aspects" (what you are proposing here and #1120) at the same time?
By the way, i also very much miss in Zabbix the possibility to see all the graphs of a single host at once (as provided by Ganglia for instance also) but that is the subject of ZBXNEXT-75 ...
AG, you misunderstood the spec. Currently when you open a simple graph, at some point when you select a longer period of time, you will see three values for an item: it's minimum, maximum and average value. If multiple items are selected, you see only the average value for each item. You'll still see the values for each selected item. I've corrected the spec a bit to make it less ambiquous.
Great, thanks!
I still think that for big groups of hosts, having all the lines in the same graph would make it very difficult to peek the abnormal ones. I think of a group of 100+ nodes, that is not that uncommon for bigger computing clusters
It would be great to have the possibility to switch to display a multitude of small identical graphs, one for each host. But perhaps that could be also a later improvement
(1) Translation string changes:
Added:
- Display stacked graph
- Display graph
- Specify some filter condition to see the values.
- Specify some filter condition to see the items.
- Item values
Removed:
- ITEMS
- Show items with name like
iivs CLOSED.
(2) Latest data and graph filters still need redesigning.
- Latest data filter re-design: RESOLVED in r48207, r48208
- History graph filter re-design: RESOLVED in r48212
jelisejev I've removed incorrect usages of NAME_DELIMITER in r48365. I've also changed the name filter label to just "Name".
jelisejev In r48385 I've changed the design of the history page filter buttons.
iivs I removed "main" class from filter button. It seemed unnecessary. See r48408
jelisejev Right, thanks! CLOSED.
(3) The mass-action dropdown and Go button should have disabled style like in Zabbix 2.2. It is a global issue related to all pages that have mass actions.
(4) Filter by one host, select all items. Then switch to filtering by another host. Note that the Go button still show us older selected items.
jelisejev RESOLVED in r48120.
iivs CLOSED.
(5) Host groups and Hosts does not resize when number of selected elements exceeds their size. Have a look at the configuration of templates (tab 'Linked templates') to see how it works nicely there.
iivs RESOLVED in r48188
alexei CLOSED
(6) Switching from Normal to Stacked changes item ordering. I think we should use natural sort order based on item description plus itemid.
jelisejev Added sorting in r48125. For stacked graphs items in the legend are drawn in reverse order, to match the order of the lines on the chart.
iivs CLOSED.
alexei I believe we should sort by item description plus visible host name. It would make legend much easier to read. REOPENED
iivs RESOLVED in r48326
jelisejev I've moved the sorting code out of the CLineGraphDraw class in r48379.
iivs CLOSED.
(7) Browser remebers selected checkboxes via cache. Select items with checkboxes, then press "Go" button to display graph. Now press back in browser. "Go" button is disabled, checkboxes are still checked.
RESOLVED in r48219
alexei It still does not work well. When I go back, there is no selected items however the "Go" button is still enabled and displays number of previously selected items.
iivs For some browsers checkbox selection was cached, due to this I tried to remove it with autocomplete=off attribute, but it turns out, if FireBug is disabled, checkboxes are unchecked, but yellow line style and "Go" button still displays previously selected objects. Similar things happen on other pages. Pages now read cookies on load and decide to delete them if it is a different page. So, for example, in hosts page check all hosts, click on items page and then press back, host checkboxes show same bug.
iivs I removed autocomplete=off and handled checkbox selection via JS. First it checks if checkboxes should be selected if cookies are present. If not, check if checkboxes are still selected from browser cache and then update row style, main checkbox and "Go" button if necessary.
RESOLVED in r48308
jelisejev I've made a minor change in r48369, please review. Otherwise good.
iivs CLOSED.
(8) Select/unselect all is quite slow even on my i7 CPU. It takes 2 seconds to select 1000 elements.
iivs I noticed this slowness on Chrome with 11 checkboxes only.
jelisejev RESOLVED in r48383.
iivs Opening page with for example 90 checkboxes and no checkboxes selected, loop executes up to 8100 times.
REOPENED.
jelisejev RESOLVED in r48412.
iivs I removed checkbox debug code in r48418, other than that we can't make it faster due to row style recoloring.
jelisejev Thanks! CLOSED.
(9) Having no items selected from filter and clicking main checkbox generates JS error inside jQuery:
TypeError: e is undefined
...rCase()},each:function(e,t,n){var r,i=0,o=e.length,a=M(e);if(n){if(a){for(;o>i;i...
in jsLoader.php?ver=2.3.4&lang=en_GB&showGuiMessaging=1 (line 4978, col 5188)
iivs Not sure how I feel about main checkbox being displayed as enabled if there are no other checkboxes in the list. For now I just fixed jQuery producing error, due to empty object being passed in function.
RESOLVED in r48323
jelisejev This fix was a bit incorrect, since it wouldn't work with multiple checkbox ranges on a single page. See my fix in r48368.
iivs CLOSED.
(10) I've reverted the CSS fix you've made in r48212. It will be fixed separately under ZBX-8662.
I've also made some changes to the CSS code in r48398, please review.
iivs CLOSED.
TESTED.
Close (8) before merging.
Available in 2.3.4 r48423.
(11) Documentation needs to be updated.
martins-v Added to documentation in:
- https://www.zabbix.com/documentation/2.4/manual/introduction/whatsnew240#ad-hoc_graphs_for_several_items
- https://www.zabbix.com/documentation/2.4/manual/config/visualisation/graphs/adhoc
Also updated:
- https://www.zabbix.com/documentation/2.4/manual/web_interface/frontend_sections/monitoring/latest_data (filter changes)
RESOLVED. Please review.
<richlv> it looks like one important thing from the spec is missing in the documentation : "to avoid displaying too many lines on the graph, only the average value for each item will be displayed."
it would probably be good to review the rest of the specification in case something else could be useful to mention.
martins-v Two details added to https://www.zabbix.com/documentation/2.4/manual/config/visualisation/graphs/adhoc - about average values displayed and triggers not being displayed. RESOLVED.
jelisejev Everything else seems to be OK.
martins-v CLOSED.
(13) Having selected first option "none", the main checkbox and button appears enabled, but with is same error described in (9).
jelisejev RESOLVED in svn://svn.zabbix.com/branches/dev/ZBXNEXT-570 r48445.
iivs CLOSED.
(14) A user that has only read-only permissions, gets the following error with selecting a filter for host in latestdata: ERROR: No permissions to referred object or it does not exist!.
jelisejev Thank you for the prompt report. Fixed in svn://svn.zabbix.com/branches/dev/ZBXNEXT-570 r48448.
iivs CLOSED.
TESTED.
Fixed in 2.3.4 r48456.
filter looks pretty bad in opera - filed as ZBX-8684
turned out to be something cached
This caused a regression - ZBX-8784.
subissues still open : 11
a regression : ZBX-8847
The current implementation is nice as can compare things without creating a graph (which may not be needed at later point).
But for many cases is totally useless.
I see at least 2 situations when this implementation is not enough and we need some graphs we can define in templates/screens:
1) If there is a need to see these graphs again and again is much easier to have the graph already made and select it to see it. Having to recreate the graph again and again is time consuming
2) If with these graphs is working non-sysadmin staff (eg: developers for checking impact of code changes). They need to to be able to see quickly the graphs without bothering with details of creating an on demand graph
So even if this implementation is nice I find there is a real demand for a complementary approach...
I fully agree with the previous comment : I really enjoy this feature, but I need it in a "more stable" way, I should't have to re-create the graph every time.
I'm not a developper, nor do I fully understand zabbix, but as a user, here is how I would like it :
The graph could be linked to the hostgroup, or to the template, and adding a host to the hostgroup/template would automagically update the graph with the key of the new host.
To display the graph, maybe you could show it in every host belonging to the hostgroup.
I know I'm not bringing any solution, just (not very clear) ideas...
@Pavels Jelisejevs I really love that approach, however, I really miss the option to pin the specific graphs you create in the Latest Data section to your favorites listed in the Dashboard. Or save it anywhere.
I mean also the Dynamic Graphs
+1
Need solution to prototype group graph
Need solution to pin specific graph made on the fly