[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-05-24 Thread ASF subversion and git services (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16489445#comment-16489445
 ] 

ASF subversion and git services commented on KNOX-1149:
---

Commit 1b053c5e11b6fca3e88c0ccfb4911e4c8bba5bed in knox's branch 
refs/heads/master from [~pzampino]
[ https://git-wip-us.apache.org/repos/asf?p=knox.git;h=1b053c5 ]

KNOX-1149 - HBase High Availability Fails with Kerberos Secured Cluster


> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Phil Zampino
>Priority: Major
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-05-21 Thread Phil Zampino (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16483245#comment-16483245
 ] 

Phil Zampino commented on KNOX-1149:


[~rkellogg] The source appears to have diverged from this patch by more than 
simply package naming. Any chance you could provide another patch to address 
this issue?

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Major
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-05-08 Thread Phil Zampino (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16467924#comment-16467924
 ] 

Phil Zampino commented on KNOX-1149:


[~rkellogg] Would you mind rebasing the patch such that the packages are 
consistent with 1.0.0+ ?

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Major
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-10 Thread larry mccay (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16320683#comment-16320683
 ] 

larry mccay commented on KNOX-1149:
---

Terrific!
I agree - that's what would make testing so extensive across all the
services.

Let's get 1.0.0 out and start on 1.1.0.

On Wed, Jan 10, 2018 at 12:06 PM, Rick Kellogg (JIRA) 



> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-10 Thread Rick Kellogg (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16320648#comment-16320648
 ] 

Rick Kellogg commented on KNOX-1149:


No objections to 1.1.0 but I wanted to have the fix available if so desired.  

Without it I believe the high availability ping operation with Kerberos will 
fail for Kafka, HBase and SOLR.  This fix definitely corrects the problem with 
HBase but I have not successfully configured the others to begin testing on 
those.  As you can imagine proper configuration is quite difficult and poorly 
documented.

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-10 Thread Larry McCay (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16320475#comment-16320475
 ] 

Larry McCay commented on KNOX-1149:
---

Hi [~rkellogg] - I think this looks like an important fix.
Unfortunately, the fact that it "should also correct issues with other Kerberos 
protected services" means two things:

1. this will require extensive testing across all services in a secure cluster
2. this will potentially change behavior for many services from 0.14.0

#1 will push out the 1.0.0 release and delay work on 1.1.0 and #2 will risk 
destabilizing 1.0.0 in a way that is different from 0.14.0.

My inclination is to leave it in 1.1.0 and get 1.0.0 out as is.
We can then immediately begin work on 1.1.0 and try and managed scope to get 
such fixes in.

Thoughts?


> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
> Fix For: 1.1.0
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-08 Thread Larry McCay (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16317606#comment-16317606
 ] 

Larry McCay commented on KNOX-1149:
---

Okay, I think that we should actually move it to Major as it is really broken.
Since there are a couple workarounds and you don't mind deferring it, I'll move 
it to 1.1.0 for now.

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-08 Thread Rick Kellogg (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16317591#comment-16317591
 ] 

Rick Kellogg commented on KNOX-1149:


I can state that under HDP 2.6, the HBase Region Servers are listed under a 
different path.  So it is fundamentally broken with HDP at least.  

If someone were to set the value explicitly in the hbase-site.xml file they 
could work around the issue.  Another option is explicitly listing the URLs 
under the service which defeats the HA feature.  No objections to deferral as I 
am testing another fix related to the ping operation with Kerberos as well.  
More to report tomorrow after I do some more testing.

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2018-01-08 Thread Larry McCay (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16317581#comment-16317581
 ] 

Larry McCay commented on KNOX-1149:
---

[~rkellogg] - we are descoping 1.0.0 to make sure it is as functionally similar 
to 0.14.0 as possible. I am curious whether your Minor priority here is 
accurate. It seems like it is pretty broken for secure clusters this way. 

Is there a workaround that I am unaware of?


> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2017-12-27 Thread Rick Kellogg (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16304650#comment-16304650
 ] 

Rick Kellogg commented on KNOX-1149:


The parent node is specified in hbase-site.xml under the zookeeper.znode.parent 
value.

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2017-12-21 Thread Rick Kellogg (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16300508#comment-16300508
 ] 

Rick Kellogg commented on KNOX-1149:


Unfortunately no idea why this is the case.  It might be configurable somewhere 
unknown to me.

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2017-12-21 Thread Larry McCay (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16300477#comment-16300477
 ] 

Larry McCay commented on KNOX-1149:
---

I'd like to understand why HBase has separate znodes for secure and unsecure 
clusters.
[~rkellogg] - do you have insight into this?

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 1.0.0, 0.14.1
>
> Attachments: KNOX-1149.patch
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

[jira] [Commented] (KNOX-1149) HBase High Availability Fails with Kerberos Secured Cluster

2017-12-19 Thread Rick Kellogg (JIRA) 

[ 
https://issues.apache.org/jira/browse/KNOX-1149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16297024#comment-16297024
 ] 

Rick Kellogg commented on KNOX-1149:


The normal path for Region Servers is "hbase-unsecure".  When running in a 
Kerberos secured environment the new path is "hbase-secure".

> HBase High Availability Fails with Kerberos Secured Cluster
> ---
>
> Key: KNOX-1149
> URL: https://issues.apache.org/jira/browse/KNOX-1149
> Project: Apache Knox
>  Issue Type: Bug
>  Components: Server
>Affects Versions: 0.13.0, 0.14.0
>Reporter: Rick Kellogg
>Assignee: Rick Kellogg
>Priority: Minor
> Fix For: 0.15.0, 1.0.0
>
>
> When HBase is run on a Kerberos secured cluster, the registration of the 
> Region Servers is stored in ZooKeeper under a different path.  The 
> HBaseZookeeperURLManager class used to support high availability in Knox 
> needs to be updated to look in both locations and then ping for availability.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)