Hi Josh,

Two quick questions.

1) What should I use instead of HDFS classloader? All examples seem to be from 
hdfs.
2) Whan 1.7.1 release is scheduled for (approx.) ?

Regards,
Roman

-----Original Message-----
From: Josh Elser [mailto:josh.el...@gmail.com]
Sent: 26 January 2016 19:01
To: user@accumulo.apache.org
Subject: Re: Accumulo and Kerberos

I would strongly recommend that you do not use the HDFS classloader. It is 
known to be very broken in what you download as 1.7.0. There are a number of 
JIRA issues about this which stem from a lack of a released commons-vfs2-2.1.

That being said, I have not done anything with running Accumulo out of HDFS 
with Kerberos enabled. AFAIK, you're in untraveled waters.

re: the renewal bug: When the ticket expires, the Accumulo service will die. 
Your options are to deploy a watchdog process that would restart the service, 
download the fix from the JIRA case and rebuild Accumulo yourself, or build 
1.7.1-SNAPSHOT from our codebase. I would recommend using 1.7.1-SNAPSHOT as it 
should be the least painful (1.7.1-SNAPSHOT now is likely to not change 
significantly from what is ultimately released as 1.7.1)

roman.drap...@baesystems.com wrote:
> Hi Josh,
>
> Yes, will do. Just in the meantime - I can see a different issue on slave 
> nodes. If I try to start in isolation (bin/start-here.sh) with or without 
> doing kinit I always see the error below.
>
> 2016-01-26 18:31:13,873 [start.Main] ERROR: Problem initializing the
> class loader java.lang.reflect.InvocationTargetException
>          at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
>          at 
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
>          at 
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
>          at java.lang.reflect.Method.invoke(Method.java:606)
>          at org.apache.accumulo.start.Main.getClassLoader(Main.java:68)
>          at org.apache.accumulo.start.Main.main(Main.java:52)
> Caused by: org.apache.commons.vfs2.FileSystemException: Could not determine 
> the type of file "hdfs://<hostname>/platform/lib/.*.jar".
>          at 
> org.apache.commons.vfs2.provider.AbstractFileObject.attach(AbstractFileObject.java:1522)
>          at 
> org.apache.commons.vfs2.provider.AbstractFileObject.getType(AbstractFileObject.java:489)
>          at 
> org.apache.accumulo.start.classloader.vfs.AccumuloVFSClassLoader.resolve(AccumuloVFSClassLoader.java:143)
>          at 
> org.apache.accumulo.start.classloader.vfs.AccumuloVFSClassLoader.resolve(AccumuloVFSClassLoader.java:121)
>          at 
> org.apache.accumulo.start.classloader.vfs.AccumuloVFSClassLoader.getClassLoader(AccumuloVFSClassLoader.java:211)
>          ... 6 more
> Caused by: org.apache.hadoop.security.AccessControlException: SIMPLE
> authentication is not enabled.  Available:[TOKEN, KERBEROS]
>
> I guess it might be different to what I observe on the master node. If I 
> don't get ticket explicitly, I get the error mentioned in the previous email. 
> However if do (and it does not matter for what user I have a ticket now - 
> whether it's accumulo, hdfs or hive) - it works. So I started to think, maybe 
> the problem related to some action (for example to vfs as per above) that 
> tries to access HDFS before doing a proper authentication with Kerberos? Any 
> ideas?
>
> Also, if we go live with 1.7.0 - what approach would you recommend for 
> renewing tickets? Does it require stopping and starting the cluster?
>
> Regards,
> Roman
>
>
>
> -----Original Message-----
> From: Josh Elser [mailto:josh.el...@gmail.com]
> Sent: 26 January 2016 18:10
> To: user@accumulo.apache.org
> Subject: Re: Accumulo and Kerberos
>
> Hi Roman,
>
> Accumulo services (TabletServer, Master, etc) all use a keytab to 
> automatically obtain a ticket from the KDC when they start up. You do not 
> need to do anything with kinit when starting Accumulo.
>
> One worry is ACCUMULO-4069[1] with all presently released versions (most 
> notably 1.7.0 which you are using). This is a bug in which services did not 
> automatically renew their ticket. We're working on a 1.7.1, but it's not out 
> yet.
>
> As for debugging your issue, take a look at the Kerberos section on debugging 
> in the user manual [2]. Take a very close look at the principal the service 
> is using to obtain the ticket and what the principal is for your keytab. A 
> good sanity check is to make sure you can `kinit` in the shell using the 
> keytab and the correct principal (rule out the keytab being incorrect).
>
> If you still get stuck, collect the output specifying 
> -Dsun.security.krb5.debug=true in accumulo-env.sh (per the instructions) and 
> try enabling log4j DEBUG on org.apache.hadoop.security.UserGroupInformation.
>
> - Josh
>
> [1] https://issues.apache.org/jira/browse/ACCUMULO-4069
> [2]
> http://accumulo.apache.org/1.7/accumulo_user_manual.html#_debugging
>
> roman.drap...@baesystems.com wrote:
>> Hi there,
>>
>> Trying to setup Accumulo 1.7 on Kerberized cluster. Only interested
>> in master/tablets to be kerberized (not end-users). Configured
>> everything as per manual:
>>
>> 1)Created principals
>>
>> 2)Generated glob keytab
>>
>> 3)Modified accumulo-site.xml providing general.kerberos.keytab and
>> general.kerberos.principal
>>
>> If I start as accumulo user I get: Caused by: GSSException: No valid
>> credentials provided (Mechanism level: Failed to find any Kerberos
>> tgt)
>>
>> However, if I give explicitly a token with kinit and keytab generated
>> above in the shell - it works as expected. To my understanding
>> Accumulo has to obtain tickets automatically? Or the idea is to write
>> a cron job and apply kinit to every tablet server per day?
>>
>> Regards,
>>
>> Roman
>>
>> Please consider the environment before printing this email. This
>> message should be regarded as confidential. If you have received this
>> email in error please notify the sender and destroy it immediately.
>> Statements of intent shall only become binding when confirmed in hard
>> copy by an authorised signatory. The contents of this email may
>> relate to dealings with other companies under the control of BAE
>> Systems Applied Intelligence Limited, details of which can be found
>> at http://www.baesystems.com/Businesses/index.htm.
> Please consider the environment before printing this email. This message 
> should be regarded as confidential. If you have received this email in error 
> please notify the sender and destroy it immediately. Statements of intent 
> shall only become binding when confirmed in hard copy by an authorised 
> signatory. The contents of this email may relate to dealings with other 
> companies under the control of BAE Systems Applied Intelligence Limited, 
> details of which can be found at 
> http://www.baesystems.com/Businesses/index.htm.
Please consider the environment before printing this email. This message should 
be regarded as confidential. If you have received this email in error please 
notify the sender and destroy it immediately. Statements of intent shall only 
become binding when confirmed in hard copy by an authorised signatory. The 
contents of this email may relate to dealings with other companies under the 
control of BAE Systems Applied Intelligence Limited, details of which can be 
found at http://www.baesystems.com/Businesses/index.htm.

Reply via email to