SNMPV3

Dani@M3T

Ok I will try on one AP.
So only NetXMS agent-, routing table- and topology-polling are disabled, everything else is enabled again. I will check and report.

Beside I also have SNMP_AUTH_FAILURE events on these nodes (since V2.0-RC2)

Dani@M3T

Hi Victor

The test AP is now stable for 18h. I only have tested with one of the affected APs because of some other APs I do not have credentials to reboot. I have to contact my customer each time.
Do you have a suspicion?

thanks
Dani

gmonk63

Dani@M3T

I think you are still experiencing this issue https://www.netxms.org/forum/configuration/snmp-credentials/ I dont think the patch fixed this particular error where even though snmpv3 credentials now work the server is still sending out some requests with public as default .. I am seeing the same thing on my test switch snmpv3 and v2 works great now but I am still seeing snmp auth failures about 100 a day but on our switches there is a threshold that can be set I think the default is 10 failures in a minute before its stops responding which i increased untill this gets fixed ....

Dani@M3T

I know the other issue. But I'm not so sure it is the same. One test node only with deactivated topology and routing table polling is stable for 24h.

Victor Kirhenshtein

So it seems that topology or routing table polls cause SNMP agent hang in access points. Most likely it's a bug in SNMP agent. We can try to narrow it further by enabling one of the polls. Topology poll usually do more requests than routing table poll.

Best regards,
Victor

Dani@M3T

Hi Victor

I made these tests:

reenabled routing table poll, waiting 20h -> SNMP agent on node still replying
reenable topology polling, waiting about 20h -> SNMP agent dead

But what can we do? Is it a problem in NetXMS or in SNMP implementation on node? If it is the second I need good arguments for there support

edit:
A later test with disabled topology poll -> SNMP agent dead again
So the problem is not fixed to topology poll

thanks
Dani

Dani@M3T

I will try it with a support case at the vendor of the APs.

Another problem on other SNMP enabled nodes:
Suddenly no SNMP data is possible to poll. But the SNMP agent on these nodes are fine (test by snmpwalk on command line).
When I restart netxmsd everything is fine again. It's also with SNMPv3.

Dani@M3T

I found a hint about this problem in the local log file of the affected node. There is a system protection function which blocks the SNMP get requests from the NetXMS server because of requests with null community string (CVE-1999-0517). In the communication settings of this NetXMS node there are valid SNMPv3 credentials of course.
So maybe this problem has the same reason as the 'SNMP auth. failures'.

tomaskir

There has been a bunch of SNMP discovery fixes in the develop branch this weekend, any chance you can build from develop and try if issues still persist?

Dani@M3T

Hi Tomas

I normaly compile from 'official' sources (at the moment V2.0-RC2) on Linux, but I have never built from develop branch.
Do you mean compiling only the server from sources from hourly Snapshot of GIT repository?

Thanks
Dani

tomaskir

Yes, you can compile from the hourly snapshot of GIT repo and test with that.
The fixes were mostly to snmp in relation to agent proxying and to snmp discovery, but maybe you will see some difference.

If none of these help (so your issue is still present after testing these newest changes), we will know yours is a totally separate issue.

The develop branch is pretty stable at the moment (I run 2 medium sized installs ~150 nodes each on it atm.), but of course be advised that it is not released code and consider backups of db/server before deploying it

Dani@M3T

Thanks Tomas.

I tried to compile hourly snapshot from git repo. But I get this error:

make[3]: Entering directory `/usr/src/netxms-2.0-dev-2015-12-14/src/libstrophe'
  CC       libstrophe_la-auth.lo
In file included from ../../include/nms_common.h:78:0,
                 from auth.c:19:
../../include/netxms-version.h:27:30: fatal error: netxms-build-tag.h: No such file or directory
 #include <netxms-build-tag.h>
                              ^
compilation terminated.
make[3]: *** [libstrophe_la-auth.lo] Error 1
make[3]: Leaving directory `/usr/src/netxms-2.0-dev-2015-12-14/src/libstrophe'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/usr/src/netxms-2.0-dev-2015-12-14/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/usr/src/netxms-2.0-dev-2015-12-14'
make: *** [all] Error 2

autoconf, flex, bison, libtool are installed.
What can I do?

Alex Kirhenshtein

Script, which generate this header require git repo, not a snapshot. It's a bug.
However, if you clone repo instead of using snapshot – it will work.
Repository url for clone is https://git.netxms.org/public/netxms.git

P.S. don't forget to switch branch to "develop" (git checkout develop)

Alex Kirhenshtein

I've fixed that, now snapshots should work as well

Dani@M3T

Thanks Alex for the fast fix

Now downloading the snapshot from download site on website, './reconf', 'configure...', 'make',... works!

Dani@M3T

The SNMP_AUTH_FAILURE events are gone!
For the other problems I need more time to test.

thanks for the fixes
Dani

gmonk63

Is the also a solution for this topic

https://www.netxms.org/forum/configuration/snmp-credentials/

im still hesitant to upgrade to 2.0 because i really dont want to have to reset over 200 switches due to the wrong community string being sent out repeatedly

Dani@M3T

In my opinion I would say yes! (11h of test)

NetXMS Forum

SNMPV3