Hide Forgot
Created attachment 520290 [details] /etc/snmp/snmpd.conf Description of problem: SNMP IP Counter do not increased when test "IP-MIB IP Statistics Table ipSystemStatsInOctets Counter Check" The test specification: http://home.englab.nay.redhat.com/~zhpeng/ct-snmpv2c-ag-1.1.0-RHEL6.2-20110822/MIB/MIB_RFC4293_ipSystemStatsInOctets.html The test log: http://home.englab.nay.redhat.com/~zhpeng/ct-snmpv2c-ag-1.1.0-RHEL6.2-20110822/MIB/9.html The test script: http://home.englab.nay.redhat.com/~zhpeng/ct-snmpv2c-ag-1.1.0-RHEL6.2-20110822/MIB/MIB_RFC4293_ipSystemStatsInOctets.seq The test ether dump file: http://home.englab.nay.redhat.com/~zhpeng/ct-snmpv2c-ag-1.1.0-RHEL6.2-20110822/MIB/9_html_Link0.pcap The net-snmp configure file: In attachment. Version-Release number of selected component (if applicable): kernel-2.6.32-191.el6.x86_64 net-snmp-5.5-34.el6.x86_64.rpm net-snmp-libs-5.5-34.el6.x86_64.rpm net-snmp-utils-5.5-34.el6.x86_64.rpm How reproducible: everytime Steps to Reproduce: 1. 2. 3. Actual results: The IP-MIB IP Statistics Table ipSystemStatsInOctets Counter did not increase. Expected results: The IP-MIB IP Statistics Table ipSystemStatsInOctets Counter should increase according to the test specification. Additional info: I downgrade packages and it passed. So it's a regression bug. downgrade packages: net-snmp-5.5-21.el6.x86_64.rpm net-snmp-libs-5.5-21.el6.x86_64.rpm net-snmp-utils-5.5-21.el6.x86_64.rpm
Created attachment 520291 [details] /usr/share/snmp/snmp.conf
Created attachment 520292 [details] /etc/rc.d/rc.local
I did not change absolutely *anything* in ipSystemStatsInOctets, icmpStatsInMsgs and ipNetToPhysicalAddress in new Net-SNMP. Isn't is just another timing issue when TAHI test suite asks for something too early? Please provide /etc/snmp/snmpd.conf, just to be sure. And is there anything in syslog? Try "grep snmp /var/log/messages" (unless you have non-default log config). BTW, where do you get net-snmp-5.5-21.el6.x86_64.rpm from? RHEL 6.0 has net-snmp-5.5-27.el6. Could you try with this one? And does net-snmp-5.5-31 from RHEL 6.1 work? (This comment fully applies also to bugs #733995 and #733997... I did not change anything there either).
> I did not change absolutely *anything* in ipSystemStatsInOctets, > icmpStatsInMsgs and ipNetToPhysicalAddress in new Net-SNMP. Isn't is just > another timing issue when TAHI test suite asks for something too early? No, it's not the TAHI scripts problem. I test it again and i found the test failed for missing package net-snmp-utils. And if net-snmp-utils installed the test will pass. I think it's a bug.What's your opinion? > Please provide /etc/snmp/snmpd.conf, just to be sure. And is there anything in > syslog? Try "grep snmp /var/log/messages" (unless you have non-default log > config). snmpd.conf is already in the attachment. And there is no log in /var/log/messages when the test failed. > (This comment fully applies also to bugs #733995 and #733997... I did not > change anything there either). bugs #733995 and #733997 failed for the same reason.
(In reply to comment #5) > > I did not change absolutely *anything* in ipSystemStatsInOctets, > > icmpStatsInMsgs and ipNetToPhysicalAddress in new Net-SNMP. Isn't is just > > another timing issue when TAHI test suite asks for something too early? > No, it's not the TAHI scripts problem. I test it again and i found the test > failed for missing package net-snmp-utils. And if net-snmp-utils installed the > test will pass. I think it's a bug.What's your opinion? Your /etc/rc.d/rc.local calls /usr/bin/snmpset, which is in net-snmp-utils package. Without these snmpsets the tests won't work. Gory details: The TAHI scripts often sends a IPv6 packet to NUT and immediately checks that appropriate counter (e.g. ipSystemStatsInOctets) has increased. But snmpd reads the counter value from kernel only once in a while (5-10 seconds?) and caches them. Therefore we use the "magic" snmpset commands in your /etc/rc.d/rc.local to turn off these caches. snmpd then always reads counter values from kernel. Without snmpset (and therefore without net-snmp-utils), the tests won't work. So, I do not see any net-snmp error here. Can we agree on closing this bug?
(In reply to comment #6) > (In reply to comment #5) > > > I did not change absolutely *anything* in ipSystemStatsInOctets, > > > icmpStatsInMsgs and ipNetToPhysicalAddress in new Net-SNMP. Isn't is just > > > another timing issue when TAHI test suite asks for something too early? > > No, it's not the TAHI scripts problem. I test it again and i found the test > > failed for missing package net-snmp-utils. And if net-snmp-utils installed the > > test will pass. I think it's a bug.What's your opinion? > > Your /etc/rc.d/rc.local calls /usr/bin/snmpset, which is in net-snmp-utils > package. Without these snmpsets the tests won't work. > > Gory details: The TAHI scripts often sends a IPv6 packet to NUT and immediately > checks that appropriate counter (e.g. ipSystemStatsInOctets) has increased. But > snmpd reads the counter value from kernel only once in a while (5-10 seconds?) > and caches them. Therefore we use the "magic" snmpset commands in your > /etc/rc.d/rc.local to turn off these caches. snmpd then always reads counter > values from kernel. Without snmpset (and therefore without net-snmp-utils), the > tests won't work. > > So, I do not see any net-snmp error here. Can we agree on closing this bug? I agree with you. Pls close the bugs, thanks.