Presumably this is related to the size of the return packet and some buffer overrun somewhere.

Adam.

Sent from BlueMail
On 10 Sep 2018, at 12:45, Attila Nagy <memetic.org@tylla.hu target=_blank>tylla_at_memetic.org@tylla.hu> wrote:
Further info about this bug:
From my testing it seems that using snmpget with lots of OID's makes the switch freeze.
More exactly: using the command line as used by Observium (quoted in my original mail) has a limit of 24, so the 25th OID causes the switch to freeze, while 24 OID's produce consistent valid output.
Interestingly, when used with the same OID (eg: ifOutOctets.
1 ifOutOctets.1 ifOutOctets.1 ..., or ifMtu.1 ifMtu.1 ifMtu.1 .... etc.) then 12 is the magic limit number.
Hmmm...

On 2018-09-10 10:49, Attila Nagy wrote:
Thanks Mike,

disabling separate_walk solved the issue for the test device. We are gonna go through all the sensitive switches and disable/test them one-by-one, and report back with the final outcome.

Is there anything we can do to further narrow down the problem?

As I mentioned we have a test hardware so we can run some tests (or we can even put it on-line if you want to get direct access to it).


Regards,
Tylla

On 2018-09-07 23:53, Mike Stupalov wrote:

Hrm,

  on all my tested TP-Link switches this feature improved polling speed.
That why separate_walk feature for this os enabled.

But such problems are certainly unacceptable, I will disable that for os.

Anyway, if you have not so many devices, goto device edit page -> Modules -> disable 'separate_walk' option.



-- 
Mike Stupalov
https://stupalov.com

-- 
Mike Stupalov
Observium Limited, http://observium.org

_______________________________________________
observium mailing list
observium@observium.org
http://postman.memetic.org/cgi-bin/mailman/listinfo/observium



observium mailing list
observium@observium.org
http://postman.memetic.org/cgi-bin/mailman/listinfo/observium