Hi,

Try these troubleshooting steps https://support.emc.com/docu93184_Isilon-Customer-Troubleshooting-Guide:-Troubleshoot-Issues-with-the-OneFS-Web-Administration-Interface.pdf?language=en_US

Thanks Josh, there's nothing wrong with the authentication service.

It's just that, it seems that disabling the SSIP to be used for accessing SSH, webgui, sftp, etc.. is one of the changes in 8.2.

Here's the pop up image:

Isilon Administration
403: Forbidden
Accessing OneFS web administration interface over a configured SmartConnect service IP address is forbidden.

Web:www.isilon.com Support:1-800-782-4362 | Worldwide: +1.508.497.7901 or http://support.emc.com

And I am able to access ssh, webgui just fine using any System zone pool IP.

But as mentioned before, I need to be using the SSIP, like I have been for years.

Thanks!

Agreed, we've just hit the same thing. We access the cluster with

https://smartconnect.clustername.example.com:8080/

and have an SSL certificate created for smartconnect.clustername.example.com. Now apparently we need a pool to access the Web UI/SSH?

An additional thought, how will the host SSH keys work if each time you attempt to connect to the cluster you get a different node, and therefore a different SSH host key?

A much worse problem: if you use a pool address, sessions break as the sessions are node local. So, as soon as you end up on a new node your session gets broken.

After upgrading to 8.2, I have an issue with the Service IP as well.  In Active Directory, the Isilon was joined to the domain using the DNS name of the Service IP (let's call that DNS name "sip").  Now, when I try to apply NTFS security to a folder (using advanced settings), or change ownership of a folder, I get "The program cannot open the required dialog box because it cannot determine whether the computer named "smartconnect" is joined to a domain"

"Smartconnect" not the name that the Isilon is joined to the domain as - it's joined using the service IP DNS name of "sip".  No DNS entries were changed - just the upgrade happened.

yep, I setup virtual Isilon 8.2 and noticed that I was no longer able to map to SMB share using \\SSIP\sharename. While I know it's not best practice to connect to specific node IP, we did have a couple of use cases where we had to point applications to specific node IP and since SSIP will move during a node going offline event ...we would use that IP to ensure connectivity for CIFS clients would not be lost. 

This is indeed a very annoying and porrly thought-out change.   While I can understand that the SSIP is intended purely to reach the delegated DNS server on port 53, the problems that this has created are legion.

The pop-up message comes from a VirtualHost definition in  /usr/local/apache24/conf/webui_httpd.conf:

ServerAdmin support@isilon.com
DocumentRoot "/usr/local/www/static"
SSLEngine on
SSLProxyEngine on
SSLCertificateFile /usr/local/apache2/conf/ssl.crt/server.crt
SSLCertificateKeyFile /usr/local/apache2/conf/ssl.key/server.key

AliasMatch ^/(.*)$ /usr/local/www/static/httpd/SSIPText.html

Require all granted

So if you connect to the SSIP and port 8080, you get the static HTML box telling you not to.  I thought that I would have to setup a separate SSL certificate for every node (!) as the certificate and key are stored locally in /usr/local/apache24/conf/ssl.key and /usr/local/apache24/conf/ssl.crt.   Swapping in new key & certificates was tedious but not hard.

However the section in webui_httpd.conf that handles connections to port 8080 on all IPs except the SSIP uses a certificate from the cache; you may have seen in KB323665 the steps to change the SSL certificate use the 'isi certificate import' command.  Again, no problem importing a certificate for every node.

BUT only one of those can have the name 'default' attached to it.  I have modified the SSLCertificateFile in each node's webui_httpd.conf to point to the respective local certificate, but on stopping and restarting Apache, that is re-written with the value from the cache of the 'default' certificate.

In effect every node MUST use the same SSL certificate, the one named as 'default'.  So you must connect using a single cluster-wide name for this to work, but you cannot do this any more with 8.2.

So far as I can see it is totally broken, as what you MUST do you CANNOT do.  Thanks Dell EMC.

So there seem to be 2 choices - (1) Make do that you can only connect securely to the 1 node that is set up as the 'default' SSL certificate (2) connect insecurely without SSL, risking your storage system passwords to eavesdropping.

In conclusion I think that the Isilon team need to come up with a new class of 'floating IP address' that if need be is only used for everything EXCEPT port 53.  We have a small cluster, but for a big cluster this would be a severe problem.

check https://community.emc.com/docs/DOC-79731 - 

OneFS 8.2: Smartconnect Service IP Address (SSIP/VIP) does not allow connections beyond DNS traffic.

The change was made because there are security concerns around allowing access to anything other than DNS to an SSIP. Is there some reason that a dynamic SmartConnect pool with a single IP isn't a suitable replacement?

I am able to create an "HA_IP" dynamic SmartConnect pool with a single IP and use it for cluster management both via that API and via the Web UI. I apologize for not posting the examples, I tried multiple times but for some reason the community web site doesn't seem to deal well with cut&paste commands and output. I can attach screen shots instead if that would be helpful.

Regards,

Tim