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.
