On 18/12/12 20:28, Chip Childers wrote:
> On Tue, Dec 18, 2012 at 9:49 AM, David Nalley <da...@gnsa.us> wrote:
>> On Tue, Dec 18, 2012 at 6:12 AM, Anshul Gangwar
>> <anshul.gang...@citrix.com> wrote:
>>> Hi,
>>> Currently CloudStack shows alerts on dashboard and can send email
>>> alerts. With this feature it will be able to send SNMP alerts to
>>> configured external SNMP Managers. Currently I am planning to send SNMP
>>> traps for the alerts which are supported in CloudStack 4.0. I am
>>> planning to use snmp4j library available under Apache License to
>>> implement this feature.
>> Fascinating - I am looking forward to seeing this!
>>
>>> - Currently planning to support up to 20 SNMP managers
> Is the proposal to tie the trap portion of the implementation to the
> messaging work that others are doing? This seems like a perfect
> candidate to be implemented as an event observer, firing traps on
> interesting events. If this is the case, then couldn't we scale the
> number of SNMP trap probe targets quite a bit higher?
>
>> This seems arbitrary. Why 20?
>>
>>> - Version 1 and 2 will be supported
>>
>> Why not v3 as well? SNMPv1 is practically deprecated, so why bother
>> with v1? v2 is a nice middle ground - v3 is where everyone wishes to
>> be if they could start from scratch.
> +1 - I can't think of a reason to have v1 support. v3 is optional
> IMO, but since we are building the feature now I don't see why it
> wouldn't be included.
>
>> So your idea here seems like it is targeting just sending traps, but
>> I'm far more interested long-term in being able to query the SNMP
>> daemon (and there likely will already be one on the host); so as you
>> are designing the MIB please do plan ahead for such things. The page
>> above is a good starting point around MIB design.
>>
> Providing an SNMP server would certainly be an interesting feature.
> It would make it much easier to implement monitoring of CloudStack
> within legacy monitoring tools (and in fact probably easier in the
> newer ones as well, but potentially less feature rich).
>
> In implementation, can we please be sure to actually create MIB files?
> Often times MIBs are skipped by software packages, frequently because
> the MIB files don't particularly help the coder implement the traps
> (the libraries don't normally use the MIB for anything). However,
> having published MIBs is exceptionally helpful when configuring a
> monitoring system.
>
>> If you need help around this, I am happy to lend what little I know
>> about the subject of SNMP, or connect you with experts in the field.
>>
> Similar to David, I'm happy to help if you would like. I've authored
> a few MIBs in my day, although all of them have been SNMPv2... so I'd
> have to learn the syntax for v3.
Do you know any open source tool to generate the MIBs?
>
>> --David
>>
