#How to identify those DNS names that have RR set for multipe LIF's ?

1 messages · Page 1 of 1 (latest)

earnest glade
#

I have several DNS names that have round robin settings across over multiple LIF's within several SVM's.
Over the time, when we create new LIF's on new joined nodes , we forgot to update the DNS entries to include those new LIF's. Further, there are multiple different DNS's names from different SVMs.
Is there anyway from ONTAP, we can identify all these LIF's on which there is DNS name been set ?

lyric sapphire
#

eh, not sure what specifics you'd be looking for, but the basic DNS options/commands
vserver services name-service dns show

you are using RR off box, correct?

earnest glade
#

No, it's not about RR off box.

indigo viper
#

Net int show -fields dns-zone,listen-for-dns-query

#

I find it way easier to use ddns

If you have the external or internal disable/delete then enable ddns

vserver services name-service dns dynamic-update modify -vserver vs1 -is-enabled true -use-secure true -vserver-fqdn vs1.abcd.com

Then do some testing

earnest glade
#

Let me explain what I need further:
Let's say I have 4 LIF's across different nodes in same vlan in a SVM. there is dns rr name setting on the top of these 4 LIF's, called "xyz.abc.com", and clients will use it to route traffic to these 4 LIF's using rond-robin.

Now, I have two new LIF's, LIF5, and LIF6 on new nodes, since we forgot about the existing LIF's have DNS rr name associating the 4 LIF's, we fogot to add LIF5 and LIF6 under the dns name. This happends to other SVM's

So, is there anyway to know these existing 4 LIF's alerady have a DNS round-roubin name associated ?

crimson nest
#

yeah, net int show -fields dns-zone should show that

#

or I'm also misunderstanding what you need 🤔

earnest glade
#

yeah, I have already ran the comand "net int show -fields dns-zone,listen-for-dns-query ", the result shows "none" and "false" on all LIF's, no differences between LIF's that have DNS RR or no DNS RR.

Any further idea's?

lyric sapphire
#

i don't think there is anything that will show what is and isn't RR, as that isn't on the netapp

ripe pasture
#

just do a dns lookup on the hostname... if all the ip addresses that come back aren't the same as what is configured for the svm, add corresponding A records...

indigo viper
#

You’re probably not doing on box dns load balance then. It’s probably off box in Active Directory

#

Just as more IPs to the name in DNS

earnest glade
#

It is not NetApp on-box DNS, but external DNS server. So, in this case, there seems no way from ONTAP to tell what are those LIF's are under a DNS name and load balanced by RR....

crimson nest
#

first you say it's not off-box DNS RR, now it's not on-box DNS, so what is it 😉

If it's on-box (i.e. DNS in ONTAP), the commands above should tell you. if it's in your external DNS, you need to do an nslookup to get all names (or check the DNS config in Windows AD or BIND or whatever your site uses)

#

there might be ways in ONTAP but it's much easier to do from your Windows / Linux client in the case of external DNS

indigo viper
#

Just do an nslookup of the name in question. If it’s set up properly it will/should tell you all four ip addresses associated with the name.
You can look on the Netapp and clearly identify those same four ip addresses

earnest glade
#

I know nslookup DNS name on Linux or Windows can list all LIF's under it. But, is there a way to tell on ONTAP side from LIF's?

The problem is, as I said, we forgot about what LIF's associated to a DNS name when we added new LIF's from new node and needed to add DNS name with RR settings.

lyric sapphire
#

there isn't anything on the netapp side that is capable of doing that without you creating some type of script to compare the info.
Aside from creating a runbook and keeping track of it

indigo viper
#

You could likely deduce by looking at interface details

System node run -node * ifstat a0a-123

It’s crude but it will show the most active.

#

Probably a cluster shell statistics command that gets similar information

#

If your DNS forward/reverse is good, “nslookup <svmname>” then “nslookup each.ip.address”

#

But come on, if you do the nslookup of the name that’s being RR, it should give you the ip addresses. Just go to ONTAP and find the addresses

We may be missing something.

sly dew
#

I think you are confusing Round Robin DNS with multiple A records in DNS. It's the job of the DNS server to provide you with all the A records of a host that is configured in the DNS servers. Depending on your dns server, it may randomize (RR in your case), but that's not part of the generic RFC spec. Some clients will decide how to handle multiple A records. Some do the 1st some take the last, some randomize it (looking like RR). Unless you are doing DDNS registration with your lif, there is no way for the DNS server to know about new lifs/IPs unless you go to your dns server and add them. I am a bit confused what issue you are trying to solve. Could you try to explain it a little more? I think I understand what you are saying but I might be missing something specific to your environment.

ancient island
#

Just set up on-box DNS round robin. It will look at the load and tell you accordingly.

indigo viper
#

That’s on box is great for small workloads. I had a customer try to use the on box dns load balancer and it ended up being a self service denial of service attack! The dns service on box could not handle the 1000’s of simultaneous requests and basically locked up for 10-20 minutes. Support said “meh” not going to fix it. Working as advertised

crimson nest
#

yeah it's a tradeoff between caching and flooding the DNS with requests. If you cache the DNS entries you don't get the "best current" LIF, and if you don't cache, every single DNS request makes it to ONTAP. I wonder if this is still a problem with recent systems though, they tend to have lots of spare CPU cycles for such things...

ancient island
crimson nest
#

Well, I can imagine in very large environments or with lots of automation (think k8s+trident) that is entirely possible, since every single mount will be a DNS request

indigo viper
#

HPC. All kicking off jobs at the same time. All needing mounts via autofs. Across 12 nodes in the cluster. Trying to spread the network load. Gave up.

#

And it was like 3000-5000 simultaneous requests that killed the cluster for everyone