MySQL Bugs: #45079: agent on FreeBSD7 does not show IP address in Meta Info on dashboard

Bug #45079	agent on FreeBSD7 does not show IP address in Meta Info on dashboard
Submitted:	25 May 2009 20:04	Modified:	14 Jan 2010 14:53
Reporter:	Carsten Segieth	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Enterprise Monitor: Agent	Severity:	S3 (Non-critical)
Version:	2.0.6.7158, 2.1.0	OS:	FreeBSD (7)
Assigned to:	Jan Kneschke	CPU Architecture:	Any
Tags:	platforms

Description:
The agents (both 32 + 64 bit) are running in a VirtualBox. In the Meta-Info on the dashboard only the localhost address is shown, but not the address that is visible from outside like it is for all other agents.

"<AGENT> --agent-run-os-tests" shows critical errors:
2009-05-21 02:08:14: (critical) network-socket.c.317: connect(127.0.0.1:3306) failed: Connection refused (61)
2009-05-21 02:08:14: (critical) sigar-test-all.c:593: sigar_net_interface_stat_get(em0) failed: 6
2009-05-21 02:08:14: (critical) sigar-test-all.c:601: sigar_net_interface_config_get(em0) failed: 6

-------------------------------------------------
59_freebsd7-x86-32_vbox-ua-1-vm1_36044  Meta Info
...	
OS          FreeBSD 7.1-RELEASE (FreeBSD 7.1-RELEASE)
CPU         CPU @ 2.93GHz
RAM         0.99 GB
Disk Space  / 13.55 GB (10.11 GB free)
IP Address  :
lo0: 127.0.0.1

<AGENT> --agent-run-os-tests
2009-05-21 02:08:14: (critical) network-socket.c.317: connect(127.0.0.1:3306) failed: Connection refused (61)
2009-05-21 02:08:14: (critical) sigar-test-all.c:593: sigar_net_interface_stat_get(em0) failed: 6
2009-05-21 02:08:14: (critical) sigar-test-all.c:601: sigar_net_interface_config_get(em0) failed: 6
sigar-test-all.c.129 (test_sigar_pid_get): 
  pid = 3342
sigar-test-all.c.111 (test_sigar_mem_get): 
  ...
sigar-test-all.c.642 (test_sigar_net_iflist_get): 
[1] 
  net.ifname = lo0, 
  net.config.name = lo0, 
  net.config.type = Local Loopback, 
  net.config.desc = lo0, 
  net.config.address = 127.0.0.1, 
  net.config.destination = 127.0.0.1, 
  net.config.hwaddr = 00:00:00:00:00:00, 
  net.config.broadcast = 0.0.0.0, 
  net.config.netmask = 255.0.0.0, 
  net.config.flags = 32841
  net.config.mtu = 16384
  net.config.metric = 0
  net.stat.rx_packets = 1851280
  net.stat.rx_bytes = 2236416917
  net.stat.rx_errors = 0
  net.stat.rx_dropped = 0
  net.stat.rx_overruns = (undef) 0
  net.stat.rx_frame = (undef) 0
  net.stat.tx_packets = 1851280
  net.stat.tx_bytes = 2236416917
  net.stat.tx_errors = 0
  net.stat.tx_dropped = (undef) 0
  net.stat.tx_overruns = (undef) 0
  net.stat.tx_collisions = 0
  net.stat.tx_carrier = (undef) 0
  net.stat.speed = 0
sigar-test-all.c.697 (0): net::connections = 

<end_of_output>

The problem exists both for the 32bit and the 64bit agent, where the above shown is nearly identical on the 64bit system.   

How to repeat:
- install and start agent on FreeBSD7

I get no failures with mysqlmonitoragent-2.1.0.1048-freebsd7-x86-32bit-installer.bin on a FreeBSD install in VMware.

Relevent output:

sigar-test-all.c.642 (test_sigar_net_iflist_get):
[0]
  net.ifname = em0,
  net.config.name = em0,
  net.config.type = Ethernet,
  net.config.desc = em0,
  net.config.address = 172.16.86.128,
  net.config.destination = (undef),
  net.config.hwaddr = 00:0C:29:D2:F2:FD,
  net.config.broadcast = 172.16.86.255,
  net.config.netmask = 255.255.255.0,
  net.config.flags = 34883
  net.config.mtu = 1500
  net.config.metric = 0
  net.stat.rx_packets = 9767
  net.stat.rx_bytes = 13133205
  net.stat.rx_errors = 0
  net.stat.rx_dropped = 0
  net.stat.rx_overruns = (undef) 0
  net.stat.rx_frame = (undef) 0
  net.stat.tx_packets = 6020
  net.stat.tx_bytes = 452739

... sigar_net_interface_stat_get(em0) failed: 6  means ENXIO 

#define ENXIO           6               /* Device not configured */

looking through the sigar code it seems it ignores:

        if (sdl->sdl_family != AF_LINK) {
            continue;
        }
        if (!((sdl->sdl_type == IFT_ETHER) ||
              (sdl->sdl_type == IFT_LOOP)))
        {
            continue; /* XXX deal w/ other weirdo interfaces */
        }

Looking http://github.com/hyperic/sigar/blob/0008c5611964d6ee1edc9938ebc3b92624a731e1/src/os/darwi... the code has been updated to handle IFT_OTHER if it has a IP address.

Carsten Segieth writes: 
testing 2.1.0.1079 the problem is still present with the same error messages "... failed: 6".

Marcos Palacios writes: 
Tested with agent build 2.1.0.1082. This agent build shows its version as "2.1.0.1079" but you can see that the install directory and the etc/mysql-monitor-agent.ini file have the right version.
It is installed in box "vbox-ua-1-vm1.mysql.com" on directory:
/home/qauser/merlin/agent/2.1.0.1082/freebsd7-x86-32bit/vbox-ua-1-vm1/

The agent can be seen in the dashboard as "59_freebsd7-x86-32_vbox-ua-1-vm1_36044" on http://qa-merlin:38080/Monitor.action

Result:
In the Meta-Info on the dashboard only the localhost address is shown, but not the address that is visible from outside like it is for all other agents.

Jan Kneschke writes: 
in src/os/darwin/darwin_sigar.c line 2562 is:

          case IFMSG_ITER_GET:
            if (strEQ(iter->name, sdl->sdl_data)) {
                iter->data.ifm = ifm;
                return SIGAR_OK;
            }

As the comment above that block says:

   /* sdl_data doesn't include a trailing \0, it is only sdl_nlen long */

which breaks the strEQ() macro which does a strcmp().

Replacing it by:

    while (ptr < end) {
       int len;
       ...
          case IFMSG_ITER_GET:
            len = strlen(iter->name);
            if (len == sdl->sdl_nlen && 0 == memcmp(iter->name, sdl->sdl_data, len)) {

fixes it.

(mail sent to hyperic)

Waiting for it to hit the master-tree so we can merge it back.

Gary Whizin writes: 
didn't make final beta release, moving to gamma

Jan Kneschke writes: 
tracked at hyperic as http://jira.hyperic.com/browse/SIGAR-137

Jan Kneschke writes: 
sigar is updated to a version that should include a fix for it.

Carsten Segieth writes: 
IP address problem solved with 2.1.1.1114, but one of the 2 FreeBSD7 VBox systems now shows 15 CPUs. But this is another problem and will be reported separate.

A note has been added to the 2.1.1 changelog: 

The IP address of the agent host on FreeBSD 7 systems would not be reported correctly.