0.8.4 Updates/Roadmap

[raX]

It has been over two months since the last release of Cacti, and 0.8.4 is just on the horizon. I am going to try and road map some things and put some thoughts on the table so everyone has a better idea of where Cacti is headed in the short term.

First of all, I'll talk about what is new in 0.8.4. As always there are bug fixes in this version, however there are several really annoying ones that have been haunting me ever since 0.8.3a was released. The first thing you will see when you login to 0.8.4, is the menu has been restructured a bit. One of the most common things that I hear from users regarding previous versions is that it is not obvious to new users how to create traffic and other host-related graphs. They are right, going through "Polling Hosts" to access this functionality does seem a bit counter-intuitive. Now, the first menu item you see is called "Create Graphs". All of the graph related functionality that used to reside under "Polling Hosts" now resides here. I hope that you all enjoy the Javascript magic that I put into this page as well. It should make creating graphs easier and prevent duplicate graphs from being created in the first place.

In addition to moving graph related features out of polling hosts, I added a little polish to this page as well. The "Associated Data Queries" box is much improved UI-wise and now allows users to run a data query in verbose mode, which is great for debugging. If the host is SNMP-enabled, you will see basic SNMP information (system, uptime, hostname) displayed on top of the page. This is a good way to see at a glance if SNMP is working for a particular host.

Now onto the big feature for 0.8.4, template import/export support. I call this a big feature for two reasons. One being that it took quite a while to code and come up with a design scalable enough to last quite a few versions of Cacti. The second reason is all of the great things this will allow Cacti to do. On a basic level, you can now setup all of your host/graph/data templates on one installation, export them to an XML file, and instantly import them into another installation of Cacti. All of the dependences, duplicates, etc are taken care of. Now imagine how cool it would be if we started collecting these XML files for all types of systems into a central repository. This in fact is my intention behind this feature. You may not see this immediately, but sometime down the road I will be creating a site that will contain templates for every host type I can get my hands on. Whether you want to graph a Juniper router or a Solaris 9 box, just grab the appropriate XML file and import it into your installation of Cacti.

As of right now, 0.8.4 is just on the horizon. In case you do not read the beta versions forum, I am asking all users to test out 0.8.4-pre before I release it. My goal is to make this release as rock solid as possible, which means hammering out the bugs before it is released.

After 0.8.4 is released, I plan to make polishing Cacti my goal for 0.8.5. This includes things like documentation (code and user), error reporting, and website stuff. I will probably also take a good hard look through the feature requests forum and add features that look useful and viable.

Any and all feedback is appreciated .

-Ian

[Deano]

I've been hooked since i found cacti. I'd been putting off migrating loads of management graphing from MRTG to RRD because I hadnt found a combined Front/Back that made life easy.

I cant wait for the templates/queries repository - It will be easier to write our own the more examples there are to look at.

I've added a couple of requests I stumbled over during my migration to the Feature request forum. I now have 400+RRAs and growing. I love the ability to present the same data in several diffeent ways without impacting the performance of the actual devices by gathering the stats multiple times (which I had to sometimes with MRTG)

We also have access to Concord Network Health for detailed historical analysis - but for sheer ease of use and instant viewing of my core network Cacti/RRD has it beat

Thanks for all your hard work on Cacti

Deano

[SeanLee]

Everything looks great. I too am looking forward to the templates/queries repository. The only thing I have not been able to do is graph *multiple* CPUs on my RH 9 boxes (via snmp). Only one graph seems to be appearing.

[raX]

Everything looks great. I too am looking forward to the templates/queries repository. The only thing I have not been able to do is graph *multiple* CPUs on my RH 9 boxes (via snmp). Only one graph seems to be appearing.

I don't think graphing multiple CPUs on a per-CPU basis works with net-snmp. But if you have two CPUs for instance, the graph will go up to 200% if both CPUs are in use.

-Ian

[SeanLee]

I don't think graphing multiple CPUs on a per-CPU basis works with net-snmp. But if you have two CPUs for instance, the graph will go up to 200% if both CPUs are in use.
-Ian

I was beginning to wonder if that's what was happening (200%)... Is there an alternative way to graph multiple CPU's (linux-redhat) that you are aware of?

[raX]

I was beginning to wonder if that's what was happening (200%)... Is there an alternative way to graph multiple CPU's (linux-redhat) that you are aware of?

Not that I know of. However if you could somehow retrieve this information from the /proc filesystem, it would be easy to add to net-snmp.

-Ian

[mike_tcis]

After 0.8.4 is released, I plan to make polishing Cacti my goal for 0.8.5. This includes things like documentation (code and user), error reporting, and website stuff. I will probably also take a good hard look through the feature requests forum and add features that look useful and viable.

Any and all feedback is appreciated .

-Ian

One very useful feature would be the ability to remove "spikes" from the graphs. We do nightly tape backups on all of our machines, which causes large (>30MB) spikes for sustained periods on our graphs. While this is OK for our internal machines, it throws off our Co-location customer graphs, especially when we use them to bill for bandwidth usage. Yes, I know I could have a totally seperate Ethernet segment for backups, but that simply isn't feasible in our shop. So, consider this a feature request

--Mike

[Guest]

After 0.8.4 is released, I plan to make polishing Cacti my goal for 0.8.5. This includes things like documentation (code and user), error reporting, and website stuff. I will probably also take a good hard look through the feature requests forum and add features that look useful and viable.

Any and all feedback is appreciated .

-Ian

One very useful feature would be the ability to remove "spikes" from the graphs. We do nightly tape backups on all of our machines, which causes large (>30MB) spikes for sustained periods on our graphs. While this is OK for our internal machines, it throws off our Co-location customer graphs, especially when we use them to bill for bandwidth usage. Yes, I know I could have a totally seperate Ethernet segment for backups, but that simply isn't feasible in our shop. So, consider this a feature request

--Mike

[TheWitness]

I would run a cron to disable the hosts prior to backups and re-enable them after. Spike removal is for short 1 interval bursts created by a host reboot and counter reset that occur on some devices (like Servers) that don't store the OID values in non-volatle memory to protect against spikes.

TheWitness