Re: Hadoop and Ceph integration issues

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



In the log file that you showing, do you see where the keyring file is
being set by Hadoop? You can find it by grepping for: "jni: conf_set"

On Mon, Sep 23, 2013 at 12:43 PM, Rolando Martins
<rolando.martins@xxxxxxxxx> wrote:
> bin/hadoop fs -ls
>
> Bad connection to FS. command aborted. exception:
>
> (no other information is thrown)
>
> ceph log:
> 2013-09-23 19:42:27.545402 7f0b58de7700 -1 monclient(hunting): ERROR:
> missing keyring, cannot use cephx for authentication
> 2013-09-23 19:42:27.545619 7f0b58de7700 20 client.-1 trim_cache size 0 max 0
> 2013-09-23 19:42:27.545733 7f0b58de7700 10 jni: ceph_mount: exit ret -2
>
> On Mon, Sep 23, 2013 at 3:39 PM, Noah Watkins <noah.watkins@xxxxxxxxxxx> wrote:
>> What happens when you run `bin/hadoop fs -ls` ? This is entirely
>> local, and a bit simpler and easier to grok.
>>
>> On Mon, Sep 23, 2013 at 12:23 PM, Rolando Martins
>> <rolando.martins@xxxxxxxxx> wrote:
>>> I am trying to start hadoop using bin/start-mapred.sh.
>>> In the HADOOP_HOME/lib, I have:
>>> lib/hadoop-cephfs.jar  lib/libcephfs.jar  lib/libcephfs_jni.so
>>> (the first I downloaded from
>>> http://ceph.com/docs/master/cephfs/hadoop/ and the other two, I copied
>>> from my system (after installing the ubuntu package for the ceph java
>>> client))
>>>
>>> I added to conf/hadoop-env.sh:
>>> export LD_LIBRARY_PATH=/hyrax/hadoop-ceph/lib
>>>
>>> I confirmed using bin/hadoop classpath that both jar are in the classpath.
>>>
>>> On Mon, Sep 23, 2013 at 3:17 PM, Noah Watkins <noah.watkins@xxxxxxxxxxx> wrote:
>>>> How are you invoking Hadoop? Also, I forgot to ask, are you using the
>>>> wrappers located in github.com/ceph/hadoop-common (or the jar linked
>>>> to on http://ceph.com/docs/master/cephfs/hadoop/)?
>>>>
>>>> On Mon, Sep 23, 2013 at 12:05 PM, Rolando Martins
>>>> <rolando.martins@xxxxxxxxx> wrote:
>>>>> Hi Noah,
>>>>> I enabled the debugging and got:
>>>>>
>>>>> 2013-09-23 18:59:34.705894 7f0b58de7700 -1 monclient(hunting): ERROR:
>>>>> missing keyring, cannot use cephx for authentication
>>>>> 2013-09-23 18:59:34.706106 7f0b58de7700 20 client.-1 trim_cache size 0 max 0
>>>>> 2013-09-23 18:59:34.706225 7f0b58de7700 10 jni: ceph_mount: exit ret -2
>>>>>
>>>>> I have the ceph.client.admin.keyring file in /etc/ceph and I tried
>>>>> with and without the
>>>>> parameter in core-site.xml. Unfortunately without success:(
>>>>>
>>>>> Thanks,
>>>>> Rolando
>>>>>
>>>>>
>>>>> <property>
>>>>>         <name>fs.ceph.impl</name>
>>>>>         <value>org.apache.hadoop.fs.ceph.CephFileSystem</value>
>>>>> </property>
>>>>>
>>>>> <property>
>>>>>         <name>fs.default.name</name>
>>>>>         <value>ceph://hyrax1:6789/</value>
>>>>> </property>
>>>>>
>>>>> <property>
>>>>>         <name>ceph.conf.file</name>
>>>>>         <value>/hyrax/hadoop-ceph/ceph/ceph.conf</value>
>>>>> </property>
>>>>>
>>>>> <property>
>>>>>         <name>ceph.root.dir</name>
>>>>>         <value>/</value>
>>>>> </property>
>>>>>  <property>
>>>>>     <name>ceph.auth.keyring</name>
>>>>>    <value>/hyrax/hadoop-ceph/ceph/ceph.client.admin.keyring</value>
>>>>> </property>
>>>>>
>>>>> On Mon, Sep 23, 2013 at 2:24 PM, Noah Watkins <noah.watkins@xxxxxxxxxxx> wrote:
>>>>>> Shoot, I thought I had it figured out :)
>>>>>>
>>>>>> There is a default admin user created when you first create your
>>>>>> cluster. After a typical install via ceph-deploy, there should be a
>>>>>> file called 'ceph.client.admin.keyring', usually sibling to ceph.conf.
>>>>>> If this is in a standard location (e.g. /etc/ceph) you shouldn't need
>>>>>> the keyring option, otherwise point 'ceph.auth.keyring' at that
>>>>>> keyring file. You shouldn't need both the keyring and the keyfile
>>>>>> options set, but it just depends on how your authentication / users
>>>>>> are all setup.
>>>>>>
>>>>>> The easiest thing to do if that doesn't solve your problem is probably
>>>>>> to turn on logging so we can see what is blowing up.
>>>>>>
>>>>>> In your ceph.conf you can add 'debug client = 20' and 'debug
>>>>>> javaclient = 20' to the client section. You may also need to set the
>>>>>> log file 'log file = /path/...'. You don't need to do this on all your
>>>>>> nodes, just one node where you get the failure.
>>>>>>
>>>>>> - Noah
>>>>>>
>>>>>>> Thanks,
>>>>>>> Rolando
>>>>>>>
>>>>>>> P.S.: I have the cephFS mounted locally, so the cluster is ok.
>>>>>>>
>>>>>>> cluster d9ca74d0-d9f4-436d-92de-762af67c6534
>>>>>>>    health HEALTH_OK
>>>>>>>    monmap e1: 9 mons at
>>>>>>> {hyrax1=10.10.10.10:6789/0,hyrax2=10.10.10.12:6789/0,hyrax3=10.10.10.15:6789/0,hyrax4=10.10.10.13:6789/0,hyrax5=10.10.10.16:6789/0,hyrax6=10.10.10.14:6789/0,hyrax7=10.10.10.18:6789/0,hyrax8=10.10.10.17:6789/0,hyrax9=10.10.10.11:6789/0},
>>>>>>> election epoch 6, quorum 0,1,2,3,4,5,6,7,8
>>>>>>> hyrax1,hyrax2,hyrax3,hyrax4,hyrax5,hyrax6,hyrax7,hyrax8,hyrax9
>>>>>>>    osdmap e30: 9 osds: 9 up, 9 in
>>>>>>>     pgmap v2457: 192 pgs: 192 active+clean; 10408 bytes data, 44312 MB
>>>>>>> used, 168 GB / 221 GB avail
>>>>>>>    mdsmap e4: 1/1/1 up {0=hyrax1=up:active}
>>>>>>>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>fs.ceph.impl</name>
>>>>>>> <value>org.apache.hadoop.fs.ceph.CephFileSystem</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>fs.default.name</name>
>>>>>>> <value>ceph://hyrax1:6789/</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>ceph.conf.file</name>
>>>>>>> <value>/hyrax/hadoop-ceph/ceph/ceph.conf</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>ceph.root.dir</name>
>>>>>>> <value>/</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>ceph.auth.keyfile</name>
>>>>>>> <value>/hyrax/hadoop-ceph/ceph/admin.secret</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> <property>
>>>>>>> <name>ceph.auth.keyring</name>
>>>>>>> <value>/hyrax/hadoop-ceph/ceph/ceph.mon.keyring</value>
>>>>>>> </property>
>>>>>>>
>>>>>>> On Mon, Sep 23, 2013 at 11:42 AM, Noah Watkins <noah.watkins@xxxxxxxxxxx> wrote:
>>>>>>>>> <property>
>>>>>>>>>         <name>ceph.root.dir</name>
>>>>>>>>>         <value>/mnt/mycephfs</value>
>>>>>>>>> </property>
>>>>>>>>
>>>>>>>> This is probably causing the issue. Is this meant to be a local mount
>>>>>>>> point? The 'ceph.root.dir' property specifies the root directory
>>>>>>>> /inside/ CephFS, and the Hadoop implementation doesn't require a local
>>>>>>>> CephFS mount--it uses a client library to interact with the file
>>>>>>>> system.
>>>>>>>>
>>>>>>>> The default value for this property is "/", so you can probably just
>>>>>>>> remove this from your config file unless your CephFS directory
>>>>>>>> structure is carved up in a special way.
>>>>>>>>
>>>>>>>>> <property>
>>>>>>>>>         <name>ceph.conf.file</name>
>>>>>>>>>         <value>/hyrax/hadoop-ceph/ceph/ceph.conf</value>
>>>>>>>>> </property>
>>>>>>>>> <property>
>>>>>>>>>         <name>ceph.auth.keyfile</name>
>>>>>>>>>         <value>/hyrax/hadoop-ceph/ceph/admin.secret</value>
>>>>>>>>> </property>
>>>>>>>>>
>>>>>>>>> <property>
>>>>>>>>>         <name>ceph.auth.keyring</name>
>>>>>>>>>         <value>/hyrax/hadoop-ceph/ceph/ceph.mon.keyring</value>
>>>>>>>>> </property>
>>>>>>>>
>>>>>>>> These files will need to be available locally on every node Hadoop
>>>>>>>> runs on. I think the error below will occur after these are loaded, so
>>>>>>>> it probably isn't your issue, though I don't recall exactly at which
>>>>>>>> point different configuration files are loaded.
>>>>>>>>
>>>>>>>>> <property>
>>>>>>>>>         <name>fs.hdfs.impl</name>
>>>>>>>>>         <value>org.apache.hadoop.fs.ceph.CephFileSystem</value>
>>>>>>>>> </property>
>>>>>>>>
>>>>>>>> I don't think this is part of the problem you are seeing, but this
>>>>>>>> 'fs.hdfs.impl' property should probably be removed. We aren't
>>>>>>>> overriding HDFS, just replacing it.
>>>>>>>>
>>>>>>>>> <property>
>>>>>>>>>         <name>ceph.mon.address</name>
>>>>>>>>>         <value>hyrax1:6789</value>
>>>>>>>>> </property>
>>>>>>>>
>>>>>>>> This was already specified in your 'fs.default.name' property. I don't
>>>>>>>> think that duplicating it is an issue, but I should probably update
>>>>>>>> the documentation to make it clear that the monitor only needs to be
>>>>>>>> listed once.
>>>>>>>>
>>>>>>>> Thanks!
>>>>>>>> Noah
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com




[Index of Archives]     [Information on CEPH]     [Linux Filesystem Development]     [Ceph Development]     [Ceph Large]     [Ceph Dev]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [xfs]


  Powered by Linux