Re: [Devmon] HP Proliant DL380

[Colin C. <col...@gm...>]

On Tue, Dec 7, 2010 at 10:26 PM, Buchan Milne
<bg...@st...> wrote:
> On Tuesday, 7 December 2010 07:01:03 Colin Coe wrote:
>> Hi all
>>
>> I'm having a problem with devmon and some HP Proliant DL380's.  The
>> problem is simply that devmon is able to extract info (raid, power,
>> temp, etc) from some but not others.
>
> This can occur when a test has some oids that may not be populated at all
> (e.g. using compaq-server on HP ProLiants with no hot spare configured end up
> with a failure polling sprDrvCntIndex, devmon gives up and doesn't poll - say
> - fans).
>
> I think I fixed this in svn with this commit:
>
> http://devmon.svn.sf.net/viewvc/devmon?view=revision&revision=156
>
> (see the explanation there)
>
> If this is not the issue, can you give more information (which tests are
> clear, details from verbose or debug logging indicating what occurs)?
>
>> To make matters less clear, on a
>> couple of the hosts where devmon works, I *cannot* snmpwalk from the
>> devmon/xymon server and on hosts where devmon doesn't work I *can*
>> snmpwalk.
>
> This makes no sense ... please confirm you are using the correct details
> devmon uses (e.g. 'cat -v hosts.db' will show records separated by '^['
> sequences). You may also want to do some packet tracing (with tcpdump or
> wireshark) if you really can't sort this out.
>
>> I've looked at the oids file and snmpwalked (for example "snmpwalk -v
>> 2c -c COMMUNITY_STR server 1.3.6.1.4.1.232.6.2.6.8.1.1") on all the
>> hosts where the "compaq;server" tests are clear, and found that I
>> could "proper" responses from most of them.
>
> Check all the oids from all tests, or apply the patch from svn, or test with
> svn trunk.
>
>> The mix of working machines includes Linux (RHEL4 & 5) and Windows.
>> The non-working machines are all Linux (RHEL4 & 5).
>>
>> All Linux nodes have the HP PSP RPMs installed (specifically
>> hp-snmp-agents) and include "dlmod cmaX /usr/lib64/libcmaX64.so" at
>> the top of /etc/snmp/snmpd.conf.
>>
>> /etc/snmp/snmpd.conf on all Linux hosts has the line: "rocommunity
>> COMMUNITY_STR monhost.company.com" where monhost.company.com is the
>> devmon/xymon server.
>>
>> The SNMP service has been restarted.
>>
>> This has me stumped.
>
>
> Regards,
> Buchan
>

Hi

I grabbed the code out of the SVN on SF and rebuilt.  It looked a lot
better but what I found was on one of the Proliants with a failed
drive but no spare, the test came back clear not red.  I made this
change:
# diff -u dm_tests.pm.orig dm_tests.pm
----
--- dm_tests.pm.orig    2010-12-09 08:07:04.000000000 +0800
+++ dm_tests.pm 2010-12-09 08:18:14.000000000 +0800
@@ -1820,9 +1820,9 @@

        # Make sure we have leaf data for our primary oid
         if(!defined $oids->{$pri}{'val'}) {
-          do_log("Missing repeater data for $pri for $test msg", 0);
-          $msg .= "&clear Missing repeater data for primary OID $pri\n";
-          $worst_color = 'clear';
+          do_log("Warning: missing repeater data for $pri for $test msg", 0);
+#          $msg .= "&clear Missing repeater data for primary OID $pri\n";
+#          $worst_color = 'clear';
           next;
         }

----
and it now comes back red although I don't know what the potential
badness of this change is (not knowing the code).

As for what I can snmpwalk and what I can't, I'm going to have to
plead insanity.  After some sleep and looking at it again I see where
I went wrong on those hosts

Thanks

CC


-- 
RHCE#805007969328369