Agent crashes when more then three managers tries to get/set/walk/getnextsimultaneously

[Kjersti Grønnevik]

Hi Frank

I have discovered that I only had the versions Agent++ 3.5.1a and Snmp++ 3.1.5
Do I have to regenerate the new code via the AgentGenerator or can I just use the new libraries directly?

Another question...
I use the Get_request and Get_next_request to update the tables within the instrumentation methods. Do I have to add sync_start() and sync_end() when modifying the table?

Thanks for fast and good reply!

Best regards,
Kjersti

-----Original Message-----
From: Frank Fock [mailto:Frank.Fock____t-online.de]
Sent: 2. september 2002 16:58
To: Kjersti Grønnevik
Subject: Re: Agent crashes when more then three managers tries to
get/set/walk/getnextsimultaneously


Hi Kjersti,

Did you see already v3.5.5b of AGENT++? Patch level (a)
fixed some multi-threading issues. These could have caused
the crashes you encountered. In particular, are you updating
tables within the instrumentation methods (get_request,
prepare_set_request, etc.) or by overwriting MibTable::update
only? If it is the last one, this could be another bug, not
yet fixed...

Thanks,
Frank


Kjersti Grønnevik wrote:
> Hi
> 
> I have the latest version of agent++ and snmp++ running on linux red hat, running multithread.
> 
> When we started testing our system - this came up:
> 
> Agent crashes when more then three managers tries to get/set/walk/getnext simultaneously 
> 
> The agent works just fine when only three managers communicates with it.
> There are no exceptions or warnings, the agent just crashes...
> 
> Does anyone have an idea why this happends? Are there a limit to how many threads that can run at the same time? If so where can I change it?
> I know I have not given much information here and I shall try to explain further if that is requested!
> 
> I would be very grateful for any suggestions or ideas I can try out!
> 
> Best regards
> Kjersti Grønnevik
>