[AGENT++] AgentX++ master agent memory usage for shared tables

Wed Oct 6 18:53:44 CEST 2010

Hi All,

I have a few questions about memory usage for the AgentX++ master agent 
(master.cpp) when an AgentX++ subagent implements shared tables (such as 
IF-MIB: if_mib.cpp and subagent.cpp).

Thanks,
Christopher Sajdak

[Question 1]: I noticed that the AgentX++ master agent seems to grow in 
memory footprint each time I start/stop the subagent repeatedly. I have 
been running the AgentX++ examples which I compiled on my RedHat linux 
machine (agent++v3.5.30, snmp++v3.2.24, agentX++v1.4.17). I suspect that 
add_row() may be involved, but not sure how the master agent would be 
growing. Any ideas? 

$ ps -F -p 5735
UID        PID  PPID  C    SZ   RSS PSR STIME TTY          TIME CMD
root      5735 27795  0 37137  4832   7 09:46 pts/17   00:00:00 ./master 
$ ps -F -p 5735
UID        PID  PPID  C    SZ   RSS PSR STIME TTY          TIME CMD
root      5735 27795  0 37201  5144   7 09:46 pts/17   00:00:00 ./master
$ ps -F -p 5735
UID        PID  PPID  C    SZ   RSS PSR STIME TTY          TIME CMD
root      5735 27795  0 37265  5396   7 09:46 pts/17   00:00:00 ./master
$ ps -F -p 5735
UID        PID  PPID  C    SZ   RSS PSR STIME TTY          TIME CMD
root      5735 27795  0 37265  5400   7 09:46 pts/17   00:00:00 ./master
$

Example files mentioned:
agentpp/agentX++/examples/master/src/master.cpp
agentpp/agentX++/examples/subagent/src/subagent.cpp
agentpp/agentX++/examples/subagent/src/if_mib.cpp

NOTE:I modified the main loop in the subagent.cpp example to allow for a 
reconnection to the master agent. This allows me to stop and restart the 
master agent as well. See the code snippet here:

   // Replace the main loop in 
agentpp/agentX++/examples/subagent/src/subagent.cpp with this:
   // Main Loop
   Request* req;
   int retries;
   do
   {
      while ((run) && (!mib->get_agentx()->quit()))
      {

         req = reqList->receive(40000);

         if (req)
         {
            mib->process_request(req);
         }
         else
         {
            mib->cleanup();
         }
      }
      mib->save_all(); // save persistent data to disk
      retries = 0;
      // Try to reconnect 10 times every 10 seconds
      while ((run) && (retries++ < 10) && (!mib->init()))
      {
         printf("Sleep for 10 sec during retry... \n");
         sleep(10);
      }
   }
   while ((run) && (retries < 10)); // stop after 10 consecutive failures

[Question 2]: I have noticed when I try to use allocate_index() or 
init_row() for a shared table with multiple indices ("1.1", "1.2", "1.3", 
"2.1", "2.2", "2.3"), I get these errors (259 and 260) on some of the 
entries. Any tips?

./agentx_pp/agentx_def.h:#define AGENTX_INDEX_ALREADY_ALLOCATED 259
./agentx_pp/agentx_def.h:#define AGENTX_DUPLICATE_REGISTRATION  263

allocate_index() ok on "1.1" and "2.2"
allocate_index() fails on "1.2", "1.3", "1.4", "2.1", "2.3", "2.4"

init_row() ok on "1.1" and "2.2"
init_row() fails on "1.2", "1.3", "1.4", "2.1", "2.3", "2.4"

(BTW, I noticed a thread: 
http://lists.agentpp.org/pipermail/agentpp/2008-January/004210.html that 
might be the same thing I am facing).

Meanwhile these would all be handled by add_row(), but I think the master 
agent is consuming extra memory in the process.

[Question 3]: When running the AgentX++ example subagent.cpp (implements 
IF-MIB), I see that there are several entries for each index. (see 
IF-MIB::ifIndex output from snmpwalk below). What is the purpose of these 
entries? Are they needed or can they be omitted somehow? 

$ snmpwalk -v3 -uunsecureUser localhost:4700 | grep IF-MIB
IF-MIB::ifNumber.0 = INTEGER: 13
IF-MIB::ifType.1 = INTEGER: softwareLoopback(24)
IF-MIB::ifType.2 = INTEGER: ethernetCsmacd(6)
IF-MIB::ifType.3 = INTEGER: ethernetCsmacd(6)

(snip)

IF-MIB::ifOutErrors.11 = Counter32: 0
IF-MIB::ifOutErrors.12 = Counter32: 0
IF-MIB::ifOutErrors.13 = Counter32: 0
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.118 = OID: IF-MIB::interfaces  <--- 
(AGENTX-MIB::agentxRegStart = .1.3.6.1.2.1.74.1.4.2.1.3)
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.131 = OID: IF-MIB::ifIndex.1     <--- 
are these entries added for each shared table index allocated?
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.138 = OID: IF-MIB::ifIndex.2 
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.139 = OID: IF-MIB::ifIndex.6
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.141 = OID: IF-MIB::ifIndex.5
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.152 = OID: IF-MIB::ifIndex.7
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.158 = OID: IF-MIB::ifIndex.10
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.192 = OID: IF-MIB::ifIndex.8
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.204 = OID: IF-MIB::ifIndex.3
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.208 = OID: IF-MIB::ifIndex.9
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.217 = OID: IF-MIB::ifIndex.11
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.220 = OID: IF-MIB::ifIndex.4
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.221 = OID: IF-MIB::ifIndex.12
SNMPv2-SMI::mib-2.74.1.4.2.1.3.2.2.223 = OID: IF-MIB::ifIndex.13

Now I stop the master agent and then restart it, I see that the 
IF-MIB::ifIndex entries are no longer there. Could these be contributing 
to master agent memory overhead?

$ snmpwalk -v3 -uunsecureUser localhost:4700 | grep IF-MIB
IF-MIB::ifNumber.0 = INTEGER: 13
IF-MIB::ifType.1 = INTEGER: softwareLoopback(24)
IF-MIB::ifType.2 = INTEGER: ethernetCsmacd(6)
IF-MIB::ifType.3 = INTEGER: ethernetCsmacd(6)

(snip)

IF-MIB::ifOutErrors.11 = Counter32: 0
IF-MIB::ifOutErrors.12 = Counter32: 0
IF-MIB::ifOutErrors.13 = Counter32: 0
SNMPv2-SMI::mib-2.74.1.4.2.1.3.1.1.1 = OID: IF-MIB::interfaces     <-- 
this time I don't see IF-MIB::ifIndex entries after master agent restart
[cjsajdak at flexnetlab_backup ~]$ 