Re: [load balancing] F5 LTM marking nodes down

From: Julian Grunnell <julian.grunnell [izzat] pipex.net>
Date: Thu Nov 08 2007 - 05:05:46 EST

Yep, agree the health monitors are nothing to do with this. There's only 2
nodes and not a lot of traffic being done at all so capturing tcpdumps is
very straight forward and been done a number of times when the problem is
seen.

J.

> -----Original Message-----
> From: lb-l-bounces@vegan.net [mailto:lb-l-bounces@vegan.net]
> On Behalf Of LeMay, John
> Sent: 07 November 2007 16:42
> To: Load Balancing Mailing List
> Subject: Re: [load balancing] F5 LTM marking nodes down
>
>
> I agree, however increasing those timeouts may hide the issue
> or as I stated possibly indicate a timing issue. The trace is
> ideal, but depending on how much traffic is going through the
> device that may be like looking for a needle in a haystack.
>
>
> --
> John LeMay
> K. Hovnanian Companies, LLC
> Systems Engineer
> Ph. 732 383-2195
> jlemay@khov.com
>
> -----Original Message-----
> From: lb-l-bounces@vegan.net [mailto:lb-l-bounces@vegan.net]
> On Behalf Of Omachonu Ogali
> Sent: Wednesday, November 07, 2007 11:22
> To: Load Balancing Mailing List
> Subject: Re: [load balancing] F5 LTM marking nodes down
>
> But if the issue was the frequency of the monitor, he would
> be seeing intermittent up/down events from both units, not
> just the standby. A network sniff is really needed to see
> exactly what is going on.
>
> oo
>
> On Nov 7, 2007 10:48 AM, LeMay, John <JLeMay@khov.com> wrote:
> >
> >
> >
> > A possible resolution, really a workaround, may be to use custom
> > monitors for http and icmp with increased timer and interval values.
> > I've done this in the past for different reasons, and it
> doesn't tell
> > you where the issue is of course. If you do go this route, F5 has
> > recommended to me in the past that the timeout should always equal
> > three times the interval plus one second (the online help
> recommends
> > the same). This may even give a clue to the root cause of the issue.
> > If you increase the timeout and interval and the issue goes away,
> chances are the issue is latency of some sort.
> >
> > If you can get this to happen at predictable times, you
> could always
> > run tcpdump on the standby box during that time period and see if
> > there are any clues in the trace.
> >
> > --
> > John LeMay
> > K. Hovnanian Companies, LLC
> > Systems Engineer
> > Ph. 732 383-2195
> > jlemay@khov.com
> >
> >
> >
> > ________________________________
> >
> > From: lb-l-bounces@vegan.net
> [mailto:lb-l-bounces@vegan.net] On Behalf
>
> > Of Julian Grunnell
> > Sent: Wednesday, November 07, 2007 10:14
> > To: lb-l@vegan.net
> > Subject: [load balancing] F5 LTM marking nodes down
> >
> >
> >
> >
> > Hi - got a query regarding alerts that a STANDBY LTM 1500 is
> > generating, we have a pair of LTM 1500's in an ACTIVE /
> STANDBY setup
> using vlangroups.
> > Version 9.3.0 is installed. The nodes behind the LTMs are Microsoft
> > 2003 IIS 6.0 servers using ACTIVE / PASSIVE NIC teaming.
> >
> > We see on a daily basis the STANDBY LTM marking nodes as
> DOWN and then
>
> > UP for both icmp and HTTP health checks, often a minute or two
> > inbetween the state changes, the ACTIVE LTM NEVER logs anything.
> >
> > At the times the alerts are seen the guys that admin the
> web servers
> > cannot see anything wrong at all - the servers are up and
> running. And
>
> > the guys that admin the network infrastructure cannot see
> any errors
> > at all from any of the switchports. What I have seen is
> that when the
> > STANDBY LTM has the node marked DOWN I can both ping the node in
> > question and make a port 80 connection from the STANDY LTM which is
> just bizarre.
> >
> > If anyone has any ideas on where the problem might lie, things that
> > would be worth checking or even better have had this problem before
> > and how to resolve it I would appreciate it!!
> >
> > Thanks - Julian.
> >
> >
> > Julian Grunnell
> > 3rd Line Technical Support
> > Pipex Communications
> >
> > Tel: 01296 300227
> > Mob: 07803 649593
> > Web: http://www.pipex.com/
> >
> > This e-mail is subject to: http://www.pipex.net/disclaimer.html
> >
> > _______________________________________________
> > lb-l mailing list
> > lb-l@vegan.net
> > http://vegan.net/mailman/listinfo/lb-l
> > Searchable Archive: http://vegan.net/lb/archive http://lbdigest.com
> > Load Balancing Digest Take the survey:
> > http://www.surveymonkey.com/s.aspx?sm=bpQsLdWxilL3kW3FyWjKvw_3d_3d
> >
> _______________________________________________
> lb-l mailing list
> lb-l@vegan.net
> http://vegan.net/mailman/listinfo/lb-l
> Searchable Archive: http://vegan.net/lb/archive
> http://lbdigest.com Load
> Balancing Digest Take the survey:
> http://www.surveymonkey.com/s.aspx?sm=bpQsLdWxilL3kW3FyWjKvw_3d_3d
> _______________________________________________
> lb-l mailing list
> lb-l@vegan.net
> http://vegan.net/mailman/listinfo/lb-l
> Searchable Archive: http://vegan.net/lb/archive
> http://lbdigest.com Load Balancing Digest
> Take the survey:
> http://www.surveymonkey.com/s.aspx?sm=bpQsLdWxilL3kW3FyWjKvw_3d_3d
>

_______________________________________________
lb-l mailing list
lb-l@vegan.net
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive: http://vegan.net/lb/archive
http://lbdigest.com Load Balancing Digest
Received on Thu Nov 8 05:25:04 2007

This archive was generated by hypermail 2.1.8 : Thu Nov 08 2007 - 05:25:05 EST