cluster da

[terry712]

used the app note from oct 2003

netware 6 cluster sp4

have a utils volume - clustered - has the public and a few bits and bobs in it. it loads the slpda -

script for load is
nss /poolactivate=UTILS
mount UTILS VOLID=253
CLUSTER CVSBIND ADD LYCAON 10.40.64.252
NUDP ADD LYCAON 10.40.64.252
add secondary ipaddress 10.40.64.252

ignore_error set slp agent ip address = follow

slpda


unload script is

unload slpda

ignore_error set slp agent ip address = first


del secondary ipaddress 10.40.64.252
CLUSTER CVSBIND DEL LYCAON 10.40.64.252
NUDP DEL LYCAON 10.40.64.252
nss /pooldeactivate=UTILS /overridetype=question

if it's on one server all is fine
if i roll it - it loads but if i do a slpda it's inactivie - if i reset it it says unknown scope

the slp.cfg are same on each
it does loopback - i know the tids say dont but i've not had an issue with this and not sure what is best

any thoughts
all seems ok

it's like a link between the physical of one to the scope but the scope is just like to the virtual so ?

 

[marvhuffaker]

Terry, did you happen to read the Oct 2003 Appnotes? They had a big writeup about Clustering SLP and other IP services. It may help you clarify some of your issues.

Marvin Huffaker MCNE, CNE
Marvin Huffaker Consulting

http://www.redjuju.com

 

[terry712]

yep that what i used to set it up

it exact as tid - worked on the test doesnt on the real

 

[terry712]

i have deleted the slpda object , scope and associated objectsand then reset up again. followed the appnote exactly

launched da on one server and it repopulated scope again - checked all servers with a display slpda and a display slp services and all ok

migrated the volume then check each server

all servers on the same segment are ok - get a positive display slpda and a positive display slp services but on remote servers - if you do a display slpda then it's ok but as soon as you do a display slp services - it deactives and you get a unknown scope

should i set this scope in the slp.cfg as well or anyone any thoughts

is it easier bringing up a secondary da somewhere and using same scope

 

[marvhuffaker]

Terry, do you have a seperate container for your SLP stuff? and do your remote servers have replicas of this container? How many DA's do you have?

Where it only seems to be a problem with your Remote stuff, and the test env. worked okay too.. I highly suspect that your remote servers are not configured correctly.

From the APP-NOTE Edirectory Design Considerations:
----------------------------------------------------
When cluster-enabling name resolution, it is strongly recommended that you incorporate the following eDirectory design considerations:

- A separate container for the DHCP information

- A separate container for the DNS information

- A separate container for the SLP information

Each of these containers must be partitioned, and every server providing any of these services must hold a replica of the corresponding container.

As always, master replicas should not be stored on a cluster. The use of a dedicated "NDS master" server is highly recommended instead.

Should you elect to use a multi-scope approach, proper DA-to-DA communication requires that all SLP DA objects be placed into the same container.

Marvin Huffaker MCNE, CNE
Marvin Huffaker Consulting

http://www.redjuju.com

 

[terry712]

the weird thing is that it's fine if it's on one side of the cluster - to me this means setup of remotes should be ok as it works on one.

i only have one da - but unless this one then i will bring up another for redundancy.

the slp objects are in a seperate container - and each cluster member has a copy of this

the remote servers dont have a replica of this - i may try to put one one a remote one and see - but again it works on one and if it was reading the ds then the slp page should reflect this - although unknown scope does more or less say thats fine but what are you talking about

cluster holds only r/w -

i am using slp.cfg as appnote says - i've always used this as multicast is disabled on routers

slp.cfg only contains the da line
discover is set to 4
the dhcp also dishes the stuff to clients and tried server on 2 - but dont like this setting as i know in past it could abend servers on boot pre sp2 or whatever


i think the plan is - 1. get both cluster server members a good clean reboot. 2. run da on one and check all ok (set slp debug to highest level ) 3. roll over and see if prob (set debug level and compare to number 2) . 4 partition scope container and pop it to each server or go with the second da

thanks - i will have a play

 

[terry712]

think i've cracked it

it's the arp cache on the routers

default setting is 4 hours so it cant ping that ip address for this period of time