SNMP and 3COM SuperStack
-
BretErickson
- Posts: 2
- Joined: Fri Dec 14, 2007 3:49 pm
SNMP and 3COM SuperStack
We've got it talking to the switch perfectly. Can you specify which port is giving the problem if you choose in/out traffic on all interfaces? Is their a variable that can be sent with the alert? Such as %interface% ??
What kind of problem? High traffic? How can we define what is "port problem" if you monitor total amount of traffic? E.g. you set HostMonitor to alert you when total amount of traffic on all interfaces is over 100Mb/sec; 4 interfaces indicates traffic like 45, 25, 35 and 30Mb/sec. Total traffic = 135Mb/sec but how can we tell which port is giving the problem? Non of them? All of them?Can you specify which port is giving the problem if you choose in/out traffic on all interfaces?
Probably you need new option, like "alert if in/out traffic on ANY interface is over N MB/sec"? Then HostMonitor should check each interface separately, if necessary set some macro variables (list of intefaces)...?
Or may be we can create new test "Dominant interface" similar to "Dominant Process" test method
http://www.ks-soft.net/hostmon.eng/news.htm#dominant700
Regards
Alex
-
BretErickson
- Posts: 2
- Joined: Fri Dec 14, 2007 3:49 pm
Thank you for your respnse. Yes, quite simply I want to know if an individual computer is using an unusually large amount of bandwidth or a port is getting a large amount of errors. We are new to this and looking for some advice. I did a good deal of reading the documentation but it didn't get very deep into this.
I'm hoping to not have to create 60 plus rules for each port.
I'm hoping to not have to create 60 plus rules for each port.
For my switches, I am using "monitor trafic on all interfaces" with alerting for "In Errors" and for "Out Errors". I would be happy to have an indication about which port originated the errors. For this, your idee about "Dominant..." ("...port" in this case) would be an enormous relief, because now I have to start a script which looks at the error counters at each port and compares to previous values of those counters - a very cumbersome procedure.
For errors, the "dominant" port is most probably the only one, which would make such an option very useful.
TIA
Alex
For errors, the "dominant" port is most probably the only one, which would make such an option very useful.
TIA
Alex
Then you should create 1 test item for each port. 60 ports - 60 tests, sorry.Yes, quite simply I want to know if an individual computer is using an unusually large amount of bandwidth or a port is getting a large amount of errors.
Yes, such test should help but it will provide infomation about sinle port only (the "worst" one). What if you already know about this port and you want to check other interfaces? May be we should add an option to exclude specific ports from the test...I would be happy to have an indication about which port originated the errors. For this, your idee about "Dominant..." ("...port" in this case) would be an enormous relief
I see another posible solution: add GUI option like "create separate test items for each interface" or improve Replicator to scan interfaces
Do you prefer to deal with smaller set of tests ("Dominant" tests)?
Regards
Alex
> Do you prefer to deal with smaller set of tests ("Dominant" tests)?
For switches it is not practical to have tests for each port, even if they would be automatically generated: although we are a relatively small company, we have nevertheless about 1000 ports, even excluding the VoIP switches.
The possibility for "monitor traffic on all interfaces" to return the "worst" port is definitely a big step forward for the "In/Out Errors", because the "worst" port would in fact be the only one - for a reasonably short (1-2-3 minutes) repeat interval.
The very best would be for such tests to return (in Reply) a list with the "worst" N ports, with N being a user-defined number. An example of Reply could be in this case something like "1/40,1/13,2/15" for a core switch with several modules, or even "1/40[2143],1/13[312],2/15[5]" with the corresponding values within [].
Further, the whole concept of "Dominant" could be generalized to "Dominant-N", as above ... but maybe I am begining to dream here ...
Regards
AlexL
For switches it is not practical to have tests for each port, even if they would be automatically generated: although we are a relatively small company, we have nevertheless about 1000 ports, even excluding the VoIP switches.
The possibility for "monitor traffic on all interfaces" to return the "worst" port is definitely a big step forward for the "In/Out Errors", because the "worst" port would in fact be the only one - for a reasonably short (1-2-3 minutes) repeat interval.
The very best would be for such tests to return (in Reply) a list with the "worst" N ports, with N being a user-defined number. An example of Reply could be in this case something like "1/40,1/13,2/15" for a core switch with several modules, or even "1/40[2143],1/13[312],2/15[5]" with the corresponding values within [].
Further, the whole concept of "Dominant" could be generalized to "Dominant-N", as above ... but maybe I am begining to dream here ...
Regards
AlexL
Thanks Jurgen.JuergenF wrote:Use SYSLOG to collect the Messages from the Switches..
My switches do not support SYSLOG, but I will check if SNMP traps could be useful.
Still, an implementation of the idea of generalized, "top N" or "dominant N", tests, would spare the user a lot of sometimes complex scripting.
What does Alex think of this?
Regards
AlexL