Graphs may select MAX instead of AVERAGE as consolidation function even if there is no item with MAX present.

[tertius1]

Since one of the 1.2.x series, some graphs will always pull the MAX CF from a *.rrd file instead of the AVERAGE CF, even if there is no graph item with a MAX CF present.

Demo graph template (with all 4 CFs referenced):

The first graph item:

This results in such a graph:

See the DEFs: only one def is created with MAX, and all LINEs are drawn with that.
Expected is this (with all 4 CFs graphed):

I tried to track down the issue and got to lib/functions.php/function generate_graph_best_cf() in line 1837.
There is this: $best_cf = '3' in line 1850. This case is visited for my graph. The step of my data source is 30, so is the parameter $ds_step, and so is $first_step.

Unfortunately, the code isn't comprehensible. I don't understand the purpose of that strange construct. If I simply deactivate that part, the graph seems to look as expected:

if ($ds_step == $first_step && isset($avail_cf_functions[1])) {
$best_cf = '3';

changed to:

if (false && $ds_step == $first_step && isset($avail_cf_functions[1])) {
$best_cf = '3';

But as I said, I don't really understand that part of the code, I just patched "something" so it works somehow better than before. I don't know if it's the correct way.

Cacti version: 1.27 (happened with 1.2.5 as well)
CentOS 7.7
rrdtool version: rrdtool-1.4.8-9.el7.x86_64

[netniV]

Basically, it looks like you disabled the ability to select the CF so it will drop to the next part. I looked at both graphs side by side and I must be tired as i'm missing the difference between them really.

[tertius1]

There are 4 graph items in the graph template. They all draw the same data source, but the 4 different Consolidation Functions within the data source: AVERAGE, MIN, MAX, LAST. In the first graph (also look at the rrdtool debugging output), only the data of the MAX consolidation function is pulled from the rrd file with the DEF: statement. All 4 lines are drawn with this data, so you see only one line.

In the second graph, where I mangled with the code, the graph generation is correct: all 4 consolidation functions are pulled from the *.rrd file, and 4 different lines are drawn. You see them as different shades of blue in the second graph screenshot. You will also see the 4 different DEF: statements in the rrdtool debugging output of the second graph screenshot, with the 4 different consolidation functions, and the 4 LINE1 drawing instructions use every 4 of them. This is how it should be.
The issue is in the cited php function - it makes cacti always pull the MAX function in this special case instead of the requested ones.

[cigamit]

This was a recent change when you are inside the first consolidation function. Since there is no aggregation, it 'should' not make a difference, but now that I think about it, the Advanced Ping template relies on the numbering of the RRDtool 'def' names, so when viewing the first RRA in the RRDfile, the template might break. Can you confirm?

[tertius1]

As far as it works on my system, this doesn't matter. For debugging I created a simple graph template that only uses the Advanced Ping data source template, not the sophisticated advanced ping graph template. I use no custom cdef's. I simply connect the test graph to one of the data sources within the existing advanced ping data source template, and it exhibits the above behavior.

Edit: (I realized later you speak of the first RRA of the first data source, not of the first data source in the *.rrd, so I redid tests and edited the original text)

In the test graph, I changed the CF from AVERAGE every other CF, so a different RRA is required. No change in behavior: Cacti continues to pull the MAX consolidation only.

Then I changed the graph template to use a different data source in the first place. No change in behavior as well.

From my system's point of view, the issue is unrelated to the advanced ping data template. It happens with every data source. It also happens regardless if it the graph pulls the first rra or any other rra. For every graph, only the MAX consolidation function is pulled into the graph, distorting every graph. To reproduce, you can take any graph, and it is immediately fixed if I re-add my hack from above, although I don't really know what it actually fixes.
What is supposed to be different between the first RRA and any other RRA?

[cigamit]

How can upload a screen show of your Data Profile Edit Screen for your Default Profile?

[tertius1]

This is the default (5 minutes collection):

And this is the data profile that is used for the RRD I detected the issue with (30 seconds polling):

[frontierliu]

I found the same problem as tertius1 and I aggree tertius1. It would make a graph DEF: statement CF always be MAX . I can not define a datasource use cf=avg/last/min.

In the Previous version, such as 1.2.5 ，the code of functions.php Line 1844 is:

if ($ds_step < 300 && isset($avail_cf_functions[1])) {

I think it was in the old version CACTI,that the minimum poller interval was 300s，and if a graph define a datasource which step<300s，no matter avg/max/min/last, always equal to MAX.
But the new version Cacti support poller interval lower than 300s, such as 60s.
So I suggest remove this code totally, direct drop to next part. should be work fine.

[cigamit]

I'm going to revert the change. In the mean time, I have to find why I thought that this was a good idea in the first place. So, should be rolled back shortly.

[cigamit]

Resolved.