Calls to Poll use fixed timeout about wapsnmp

[cdevr]

This is a complex question that has more to do with operational responsibility than with software design. The short answer is that you need to decide whether users of your tool need to be able to change the SNMP timeout (and max. retries probably).

If your operational team (the people who keep your application running) is mature enough to have valid and good opinions on this, and therefore can be trusted to handle these values responsibly (e.g. do they understand why setting it to 1h timeout and 1000 retries would be bad), then I'd say make these values configurable.

You should keep in mind that these values can change the time you need to query a router can change a lot when you change these values. If you walk a malfunctioning router for 10000+ values and set the values large, then you could find it takes hours and hours to get the information. If you set the timeout really low, then you could overload the router you're querying. So this is not a trivial issue.

[virtuallynathan]

Ah, yes, that makes quite a bit of sense. I guess i'll make a fork and allow these values to be tweaked. In my case i'm only (currently) polling 10-15 OIDs from a device that is happy to respond quickly.

Thanks!

[britcey]

Can we revisit this? I'm dealing with a bunch of equipment on the other side of the planet, so 500ms doesn't give me a lot of wiggle room. I'd prefer not to fork (this works great, otherwise, btw).

I'd assumed the 'timeout' argument to NewWapSNMP was for more than just the initial connect; having that also used for the calls to poll is what I was expecting. I'll be happy to submit a pull request replacing '500*time.Millisecond' with 'w.timeout', if that helps.

By way of comparison, the net-snmp tools default to 1 second timeout between requests (and 5 retries), both of which are configurable.

Thanks.

[cdevr]

I've had a good think about this, and I think you're right. The values should be configurable. So I've made the patch and applied it.

Commit is 2d8130a

Cheers,

Christophe