Aggregate issues after upgrade

[mppare]

I just upgraded from 0.8.8b to 1.1.38 and everything seems to be working ok. However, anytime I view an aggregate graphs, the graphs I view will subsequently experience a blip for about 25 minutes. Otherwise, graphs are generated OK. Over the weekend, I logged in to check aggregate graphs and each time I logged in - there was a corresponding gap. Otherwise - the graphs all updated no problem.

When I dug further into it this morning, I started digging into the source RRDs in one of my aggregate graphs - and confirmed that there were corresponding NaNs in the RRD. When I viewed the device (a switch), all interface graphs on that switch were updating OK (no gaps) EXCEPT for the one interface that was included in the aggregate.

Sure enough - all day as I've been working on this issue, any aggregate graphs I've touched (and the source graphs themselves) have experienced blips. One set of aggregate graphs that I did not touch/view over the weekend did not have any gaps over that period - but has gaps today each time I checked it.

I'm going to create a new aggregate from scratch, I'm also going to create one in my test environment (same Cacti version). If they work after re-creating from scratch I guess I'll call it a day (there are not that many).

Any thoughts?

[mppare]

I think I found the problem, and it has something to do with boost. I disabled boost, and it appears that now my graphs are populating correctly, even with me in the web UI browsing/viewing graphs.

If I have time I might test in my test environment (which is a base/clean install of cacti 1.1.38). I'm wondering if something (a setting or something) that migrated over from the old environment broke this functionality. But, I don't need graphs to be produced on-demand, so, it's not really a big deal.

[mppare]

Yep, it seems like a possible permissions issue related to boost that was inherited from the old server. On the old box, it looks like everything was using 'apache' to read/write the RRD files. New system defaulted to 'cacti'. I *think* this was the issue, but, since we don't really need "real time" graphs (5 min updates are OK), disabling was the quickest solution.

I used to be an mrtg admin and when I managed that, no one cared if data was up to 5-10 min old.

And, the actual real time feature works just fine.