Hi,
Long time reader, first time poster here. I'm currently running Cacti 0.8.8b, SMNP version 5.5, RRDTool 1.4.x on a Linux platform. I would like to change a handful of graphs to have the step to be 4 seconds. Yes I know that's really low, but I'm trying to catch bursts on a high speed system. I am able to change the data template, but I don't want 200 graphs with the step of 4. And since I have been running this tool a while at work, I would like to keep the historical data with in the rrd file.
Any help is appreciated.
Thanks!
Manual modification of the STEP value
Moderators: Developers, Moderators
Re: Manual modification of the STEP value
are you sure you actually CAN do that?... I fairly new here so I might be mistaken. At far as I understand a "step" is a poll, so your poller interval would have to be 4 sec, which I don't think can be done. IF I'm right, then your minimun would be 10 sec poller interval, but that would affect your entire set up and it might not work depending on your poller method and how many devices you're polling (I'm polling 52 hosts with Spine and it takes 9 sec).
Code: Select all
Corporate Cacti
CactiEZ 0.7 (upgraded to 0.0.8h) - Spine 0.8.8h - RRDTool 1.4.7 - Apache 2.2.15 - Settings 0.71 - Maint 0.3 - Thold0.5 - Errorimage 0.2 - Webmin 0.1
Watermark 0.2 - Monitor 1.3 - Realtime 0.5 - Nectar 0.35a - Fix64bit 0.4 - Autom8 0.36 - Quicktree 0.2Code: Select all
Business Cacti
CactiEZ 0.7 - Spine 0.8.8b - RRDTool 1.3.8 - Apache 2.2.15 - Settings 0.71 - Maint 0.3 - Thold0.5 - Errorimage 0.2 - Jqueryskin 0.1
Webmin 0.1 - Watermark 0.2 - Monitor 1.3 - Realtime 0.5 - Nectar 0.35a - Fix64bit 0.4 - Autom8 0.36 - Quicktree 0.2 - Syslog 1.22Re: Manual modification of the STEP value
OK, just a few things
Dropping to 4 secs steps means you will definitely lose your historical data, as it isn't based on 4 sec intervals (I assume its 60 sec or 5 minute now). Your entire RRD file is now wrong
You would need completely different RRAs created, as some of the current ones when applied at 4 secs, would mean you are retaining data for < an hour.
Just changing the RRD File isn't going to help, you would have to change the Data Template, which would affect any new graphs (not old ones). Changing the data template shouldn't affect any current graphs (unless they are recreated) with means it wouldn't also change the one you want.
In order to monitor at 4 seconds, your poller interval would have to completely change, which is going to affect everything else (4 sec poller intervals? I really don't even want to see that attempted in a lab). A single host timeout could completely throw your polling schedule over the time limit.
Ideally, what you would do...
Don't touch the current graph at all. This stops you from messing it up (and its historical data) and every other graph
Create a clone of the current data template, name it something else.
Create new RRAs for the shorter step
Assign them to the new data template
Uncheck the Data Source Active (you aren't going to want the poller to update this one)
Create a script that either runs as a daemon or via cron (cron doesn't normally do sub-minute executions) to pull the data every 4 seconds and update the RRD manually
Whether it will ever actually work, who knows... it can in theory though.
Dropping to 4 secs steps means you will definitely lose your historical data, as it isn't based on 4 sec intervals (I assume its 60 sec or 5 minute now). Your entire RRD file is now wrong
You would need completely different RRAs created, as some of the current ones when applied at 4 secs, would mean you are retaining data for < an hour.
Just changing the RRD File isn't going to help, you would have to change the Data Template, which would affect any new graphs (not old ones). Changing the data template shouldn't affect any current graphs (unless they are recreated) with means it wouldn't also change the one you want.
In order to monitor at 4 seconds, your poller interval would have to completely change, which is going to affect everything else (4 sec poller intervals? I really don't even want to see that attempted in a lab). A single host timeout could completely throw your polling schedule over the time limit.
Ideally, what you would do...
Don't touch the current graph at all. This stops you from messing it up (and its historical data) and every other graph
Create a clone of the current data template, name it something else.
Create new RRAs for the shorter step
Assign them to the new data template
Uncheck the Data Source Active (you aren't going to want the poller to update this one)
Create a script that either runs as a daemon or via cron (cron doesn't normally do sub-minute executions) to pull the data every 4 seconds and update the RRD manually
Whether it will ever actually work, who knows... it can in theory though.
Who is online
Users browsing this forum: No registered users and 1 guest