[collectd] CollectD, Collection4 (C4), and General Questions

[David Halko]

Hello List,

I have been following this software for years, did some basic
configuraiton in the past, but I am thinking about getting serious
about it.

The multicast send from a single collector to multiple receivers is
one of the most innovative features I have seen!
It is great to see that thresholding able to be sent to syslog - which
means collectd can be integrated with simplicity (really wish SNMP
Traps were supported.)

I have a couple of key questions concerning CollectD (SNMP plug-is
mostly required)
- I remember there being a top-end number of devices that a CollectD
instance could collect against... (Was it open file handles? Can
CollectD go up to 4096 yet?)
- Is there any known issues with SNMPv3 support?
- Can strings be collected daily (i.e. firmware, sysdescr, interface
names, etc.) and stored so we can check for changes?
- Is thresholding on string changes be done? (i.e. firmware update
noted from a string value change from last poll, notification that a
reboot caused interfaces to scramble from last poll, etc?)
- Is thresholding on sysuptime reset available? (i.e. sysuptime ticks
becomes lower than it was before, but some type of logic check to
ignore flip over?)

I have some general question concerning C4 and usability.
There is not much activity (updates) on this package, nor much
advertizing of C4 on the main CollectD page, but it really seems to
fit a lot of needs.
- is C4 stable for produciton work?
- are tabular reports possible (i.e. showing interfaces on a device,
firmwares across many devices, last reboot time across many devices,
etc?)
- if collecting on databases and routers, is there a way to ensure the
two groups are isolated from one another when they try to view the
RRD's?
- if collecting on databases and routers, where the two groups are
isolated from each other, is there a single-view where the admin (i.e.
me) can see it all?

My need is not so much fast collecting, but I have some HORRIBLY
CHEAP, SLOW, HIGH LATENCY, and UNRELIABLE cellular devices (~2500) and
I would like to start assembling some graphs (low bandwidth, so
occasional SNMPv1 collections ~15 minute intervals.) I also have a few
servers & databases on those servers I would like to SNMPv3 poll.
Obviously, the two groups are uninterested in each other's data and I
have a need to enforce separation.

I would appreciate some thoughts!

Thanks!