[Devmon] "devmon tests go purple" issue

[Nathan H. <na...@ma...>]

On 19/12/2007, at 3:39 AM, xbgmsharp wrote:
>
> I still say that i have not this bug on the Devmon 3 beta 3 or on my  
> wiki which is avalaible on svn trunk.
>
> -try and understand and/or resolve the "devmon tests go purple" issue.

I have some information regarding this bug.

Most switches support the CPUTotal5min oid which devmon happily  
displays in the cpu column. I have a 2900 switch running an older IOS  
which does not return any value for CPUTotal5min. When I use the 2900  
template on this old switch I get clear and/or purple messages for  
if_dsc, if_stat, if_load and if_err in addition to a clear and/or  
purple message for the cpu column.

If I modify the template to remove the CPUTotal5min oid from the cpu/ 
oids file, and remove the matching name from the cpu/message file,  
then all the other columns go green.

Similarly I have a number of HP servers with hpasm. I'm using the hp- 
server template to monitor power, fans, disks, raid, etc. I also have  
an older HP Windows 2000 server which doesn't support one of the oids  
- from memory I think it was resmem - and I get purple alerts for  
power, fans, disks, raid, etc. I had to remove resmem from the  
template before the other columns would go green.

Short summary:  If any single oid fails to retrieve then devmon stops  
collecting the rest of the oids in the template.

I've confirmed this hypothesis with with -f -p -vvvvvvvvvvv. Devmon  
says something about "too many failures" and "skipping device".  
Perhaps devmon should only skip that single column and still test the  
others.