HDP3.0: knox fails to dispatch webhdfs request due to whitelist validation

[#1]

Hi,

I am using HDP3.0 and ambari 2.7 blueprint. webhdfs via knox failed due to:

2018-08-21 19:26:33,035 ERROR knox.gateway (GatewayDispatchFilter.java:isDispatchAllowed(155)) - The dispatch to http://myhost.com:50070/webhdfs/v1/user was disallowed because it fails the dispatch whitelist validation. See documentation for dispatch whitelisting.

I have verified webhdfs without knox works:

curl -vvv http://myhost.com:50070/webhdfs/v1/user/?op=LISTSTATUS

Also, ambari, zeppelin and ranger UI work fine via knox.

The knox settings are:

gateway.dispatch.whitelist: DEFAULT

gateway.dispatch.whitelist.services: DATANODE,HBASEUI,HDFSUI,JOBHISTORYUI,NODEUI,RESOURCEMANAGER,WEBHBASE,WEBHDFS,YARNUI

webhdfs via knox worked for me on HDP2.6. Any idea? Appreciate any help.

[#2]

Looking at the documentation (https://knox.apache.org/books/knox-1-1-0/user-guide.html#Gateway+Server+Configuration), try removing

gateway.dispatch.whitelist: DEFAULT

property.

[#3 · Accepted solution]

None of Ambari, Zeppelin, or RangerUI are affected by this whitelisting.

Can you see the default whitelist in gateway.log?

It should say something like

Applying a derived dispatch whitelist because none is configured in gateway-site: xxxxxxx

[#4]

Thanks guys. I got the whitelist filter as mentioned by @Phil Zampino and updated it as my need. Then knox allowed my requests.

[#5]

what did you write in

gateway.dispatch.whitelist

???

[#6]

@Lian Jiang Can you explain why the default whitelist was not working for your deployment?

[#7]

The domain name used by hadoop hosts and the one used by the load balancer are different. DEFAULT setting will use the load balancer's domain to construct whitelist filter. I need to update the whitelist filter to use hadoop hosts' domain name instead. Hope this helps.

[#8]

Thank you for following up. That's what I suspected, and it's good to document it here for future reference.

[#9]

Hi @Phil Zampino

i have similar problem. finally what do you change ?

[#10]

@abbas mohammadnejad You have to explicitly set the gateway.dispatch.whitelist property in gateway-site.xml, such that the pattern will match the endpoint address.

[#11]

I have fill that property with

gateway.dispatch.whitelist --> ^https?:\/\/(my.domain.com|127\.0\.0\.1|0:0:0:0:0:0:0:1|::1):[0-9].*$

But i have all the time the same error

ERROR knox.gateway (GatewayDispatchFilter.java:isDispatchAllowed(155)) - The dispatch to http://my.domain.com:50070/webhdfs/v1/ was disallowed because it fails the dispatch whitelist validation. See documentation for dispatch whitelisting.

Thanks

[#12]

I ran into a similar issue to this as well. It looks like when it derives the RegEX based off the DEFAULT. We found that it used the URL from the Provider we used setting up the SSO client we had typed in just the Hostname for the Knox Server. It then resulted in the following Default white list:

INFO  knox.gateway (WhitelistUtils.java:getDispatchWhitelist(63)) - Applying a derived dispatch whitelist because none is configured in gateway-site: ^/.*$;^https?://ambari30l:[0-9]+/?.*$

After redoing the SSO setup with the FQDN for the host it resolved the "DEFAULT" Lookup and was able to find the hosts properly.

INFO  knox.gateway (WhitelistUtils.java:getDispatchWhitelist(63)) - Applying a derived dispatch whitelist because none is configured in gateway-site: ^/.*$;^https?://ambari30l.mydomain.com:[0-9]+/?.*$

Hopefully this help others in troubleshooting this issue in the future.

[#13]

Ssumers,

Can you please share the value you added in gateway.dispatch.whitelist?