haad/net-snmp

net-snmp repository with ocfs & ocfs2 support.

net-snmp's Issues

getNext can not return correct value

Hi ALL,
I have question on table handler.If I get GetNext Request OID is 1.3.6.1.4.1.39604.2.1.1.1,I fill the result in netsnmp_table_build_result(); I think the next OID should be 1.3.6.1.4.1.39604.2.1.2.1,but It is always 1.3.6.1.4.1.39604.2.1.5.0.0,what is wrong? I don't know how to fill netsnmp_table_request_info *table_info , I just fill index value, The table index is OCTE_String. Thanks a lot

How to properly use async API for multi-threading

I'm trying to write a ruby binding for this library, which only relies on the async API (because the C gem doesn't it's in ruby and does not release the global interpreter lock).

I'm using FFI for accessing C library internal structures.

The steps I do are:

• Initialize a C SNMP Session and keep the signature.
• Get the socket fileno and use it to create a ruby IO object (1. this sometimes breaks).
• I create the PDUs and send them using snmp_sess_async_send (2. if there's a lot of congestion, it breaks often).
• Use my ruby IO object to use the ruby select API call (releases the thread while waiting for work to be done) (3. sometimes hangs).
• Use snmp_sess_read, this will trigger the callback, load the response pdu in a known location from which I retrieve it, and I proceed with my life (4. this sometimes doesn't have anything there waiting).

Quite a bumpy ride. I'll proceed to explain each failure in detail:

1. Sometimes I get Errno::EBADF (the fd was marked as bad by the OS). In which circumstances I can know this? Does netsnmp provide some API to check?
2. In an example where I have 100 batches of 4 agents, each batch running on a separate thread, the async send returns 0 more often than I'd like to admit. How can I keep this under control?
3. The official API tells me to use snmp_sess_select_info to get the number of descriptors to read. So, I don't know if I'm doing it properly by waiting on events on the socket fd from the session. This seems a completely different use case though.
4. This one is also interesting. Is the callback triggered in a background thread? Because I see it being triggered, but when it comes to handling the response, it's just not there.

Could you provide me with a way of patching these issues and make this workable for ruby or any other binding-able VM? I saw that it was done for python, would be cool to achieve this for ruby.

Bug with special comment

The comment -- ----------------------------------- in .my file makes the module invisible.