FAQ
Hi,

I have a (unusual?) circumstance requiring that I run HBase against a
Zookeeper cluster which is not managed by CM. The HBase Client and Service
configuration safety valves are great in that they allow me to do this.
However, the override to the client configuration is not being pushed out.
I can log into the HBase Master and I see the
/etc/hbase/conf.cloudera.hbase1/hbase-site.xml contains a value for
hbase.zookeeper.quorum which is only the dummy zookeeper1 service I was
forced to install but do not run. zktop shows the RegionServers
communicating with my external zk cluster and everything reports cleanly on
the master's status page. However, hbase shell from the master node does
not see the correct configuration. I believe CM is managing the
configuration deployment because of the following directory structure:

$ ls -l /etc/hbase/
total 8
lrwxrwxrwx 1 root root 28 2012-06-27 03:48 conf ->
/etc/alternatives/hbase-conf
drwxr-xr-x 2 root root 4096 2012-06-27 03:52 conf.cloudera.hbase1
drwxr-xr-x 2 root root 4096 2012-06-27 00:57 conf.dist
$ ls -l /etc/alternatives/hbase-conf
lrwxrwxrwx 1 root root 31 2012-06-27 03:48 /etc/alternatives/hbase-conf ->
/etc/hbase/conf.cloudera.hbase1

Is there some other lever I need to pull, or is this a bug in client
configuration deployment?

Thanks,
-n

Search Discussions

  • Vikram Srivastava at Jun 27, 2012 at 4:25 am
    Can you paste the same configs in "Hbase Client Configuration Safety Valve"
    under "Advanced" configurations on Hbase config page and then deploy client
    configuration again for Hbase using the link from "Actions" menu?
    On Tue, Jun 26, 2012 at 9:04 PM, Nick Dimiduk wrote:

    Hi,

    I have a (unusual?) circumstance requiring that I run HBase against a
    Zookeeper cluster which is not managed by CM. The HBase Client and Service
    configuration safety valves are great in that they allow me to do this.
    However, the override to the client configuration is not being pushed out.
    I can log into the HBase Master and I see the
    /etc/hbase/conf.cloudera.hbase1/hbase-site.xml contains a value for
    hbase.zookeeper.quorum which is only the dummy zookeeper1 service I was
    forced to install but do not run. zktop shows the RegionServers
    communicating with my external zk cluster and everything reports cleanly on
    the master's status page. However, hbase shell from the master node does
    not see the correct configuration. I believe CM is managing the
    configuration deployment because of the following directory structure:

    $ ls -l /etc/hbase/
    total 8
    lrwxrwxrwx 1 root root 28 2012-06-27 03:48 conf ->
    /etc/alternatives/hbase-conf
    drwxr-xr-x 2 root root 4096 2012-06-27 03:52 conf.cloudera.hbase1
    drwxr-xr-x 2 root root 4096 2012-06-27 00:57 conf.dist
    $ ls -l /etc/alternatives/hbase-conf
    lrwxrwxrwx 1 root root 31 2012-06-27 03:48 /etc/alternatives/hbase-conf ->
    /etc/hbase/conf.cloudera.hbase1

    Is there some other lever I need to pull, or is this a bug in client
    configuration deployment?

    Thanks,
    -n
  • Nick Dimiduk at Jun 27, 2012 at 7:40 am
    That's precisely what I've done and precisely what isn't working.
    Hence the post :)

    Thanks,
    -n
    On Jun 26, 9:25 pm, Vikram Srivastava wrote:
    Can you paste the same configs in "Hbase Client Configuration Safety Valve"
    under "Advanced" configurations on Hbase config page and then deploy client
    configuration again for Hbase using the link from "Actions" menu?






    On Tue, Jun 26, 2012 at 9:04 PM, Nick Dimiduk wrote:
    Hi,
    I have a (unusual?) circumstance requiring that I run HBase against a
    Zookeeper cluster which is not managed by CM. The HBase Client and Service
    configuration safety valves are great in that they allow me to do this.
    However, the override to the client configuration is not being pushed out.
    I can log into the HBase Master and I see the
    /etc/hbase/conf.cloudera.hbase1/hbase-site.xml contains a value for
    hbase.zookeeper.quorum which is only the dummy zookeeper1 service I was
    forced to install but do not run. zktop shows the RegionServers
    communicating with my external zk cluster and everything reports cleanly on
    the master's status page. However, hbase shell from the master node does
    not see the correct configuration. I believe CM is managing the
    configuration deployment because of the following directory structure:
    $ ls -l /etc/hbase/
    total 8
    lrwxrwxrwx 1 root root   28 2012-06-27 03:48 conf ->
    /etc/alternatives/hbase-conf
    drwxr-xr-x 2 root root 4096 2012-06-27 03:52 conf.cloudera.hbase1
    drwxr-xr-x 2 root root 4096 2012-06-27 00:57 conf.dist
    $ ls -l /etc/alternatives/hbase-conf
    lrwxrwxrwx 1 root root 31 2012-06-27 03:48 /etc/alternatives/hbase-conf ->
    /etc/hbase/conf.cloudera.hbase1
    Is there some other lever I need to pull, or is this a bug in client
    configuration deployment?
    Thanks,
    -n
  • Philip Langdale at Jun 27, 2012 at 4:59 pm
    Hi Nick,

    I'm afraid this is a real bug (it works for HDFS, etc). You'll have to
    modify the deployed
    files to make the quorum match up. You can copy the directory and add a
    separate
    alternatives entry for it so that it won't get clobbered by CM if you
    redeploy.

    Also, may I ask why you aren't managing your ZK? It is unusual and not a
    supported
    configuration even though, as you say, it can be made to work.

    --phil


    On 27 June 2012 00:40, Nick Dimiduk wrote:

    That's precisely what I've done and precisely what isn't working.
    Hence the post :)

    Thanks,
    -n
    On Jun 26, 9:25 pm, Vikram Srivastava wrote:
    Can you paste the same configs in "Hbase Client Configuration Safety Valve"
    under "Advanced" configurations on Hbase config page and then deploy client
    configuration again for Hbase using the link from "Actions" menu?






    On Tue, Jun 26, 2012 at 9:04 PM, Nick Dimiduk wrote:
    Hi,
    I have a (unusual?) circumstance requiring that I run HBase against a
    Zookeeper cluster which is not managed by CM. The HBase Client and
    Service
    configuration safety valves are great in that they allow me to do this.
    However, the override to the client configuration is not being pushed
    out.
    I can log into the HBase Master and I see the
    /etc/hbase/conf.cloudera.hbase1/hbase-site.xml contains a value for
    hbase.zookeeper.quorum which is only the dummy zookeeper1 service I was
    forced to install but do not run. zktop shows the RegionServers
    communicating with my external zk cluster and everything reports
    cleanly on
    the master's status page. However, hbase shell from the master node
    does
    not see the correct configuration. I believe CM is managing the
    configuration deployment because of the following directory structure:
    $ ls -l /etc/hbase/
    total 8
    lrwxrwxrwx 1 root root 28 2012-06-27 03:48 conf ->
    /etc/alternatives/hbase-conf
    drwxr-xr-x 2 root root 4096 2012-06-27 03:52 conf.cloudera.hbase1
    drwxr-xr-x 2 root root 4096 2012-06-27 00:57 conf.dist
    $ ls -l /etc/alternatives/hbase-conf
    lrwxrwxrwx 1 root root 31 2012-06-27 03:48
    /etc/alternatives/hbase-conf ->
    /etc/hbase/conf.cloudera.hbase1
    Is there some other lever I need to pull, or is this a bug in client
    configuration deployment?
    Thanks,
    -n
  • Nick Dimiduk at Jun 28, 2012 at 2:01 am

    On Wed, Jun 27, 2012 at 9:59 AM, Philip Langdale wrote:
    I'm afraid this is a real bug (it works for HDFS, etc). You'll have to
    modify the deployed
    files to make the quorum match up. You can copy the directory and add a
    separate
    alternatives entry for it so that it won't get clobbered by CM if you
    redeploy.
    Ah. Good to know. Glad I'm not crazy.

    Also, may I ask why you aren't managing your ZK? It is unusual and not a
    supported
    configuration even though, as you say, it can be made to work.
    As I expected. The short answer is I already have a ZK available to me.
    Using it allows me to 1. not manage it myself and 2. reuse a nice bit of
    existing infrastructure around service discovery in my clients.

    Thanks for the quick response.
    -n

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupscm-users @
categorieshadoop
postedJun 27, '12 at 4:04a
activeJun 28, '12 at 2:01a
posts5
users3
websitecloudera.com
irc#hadoop

People

Translate

site design / logo © 2022 Grokbase