You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@accumulo.apache.org by Calvin Feder <ca...@argyledata.com> on 2015/01/17 01:29:26 UTC
Moving HDFS name node to another host in accumulo 1.6
We need to move the HDFS name node of our cluster to a new host (the old host is experiencing hardware issues and may die at any time). We need to keep the current accumulo data. We are trying to do it in a test environment and running into numerous issues. We stopped the cluster, moved the name node to a new host, and restarted the cluster. We changed the entry in zookeeper for the accumulo root tablet to the new location, and got accumulo to start. We see all of the tables we expected to see in accumulo, and they show up in the monitor as "online", but the monitor shows that no tablets or entries exist in those tables, and obviously we cannot scan them.
This is an example of what we are seeing:
[cid:image002.png@01D031A9.978AE280]
Has anyone tried to do something similar? What are we missing?
Thank you,
___________________________________________
Calvin Feder | QA Lead | Argyle Data, Inc.
[cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7]
________________________________
This message is for the designated recipient only and may contain privileged, proprietary, or otherwise confidential information. If you have received it in error, please notify the sender immediately and delete the original. Any other use of the e-mail by you is prohibited. Thank you in advance for your cooperation.
________________________________
Re: Moving HDFS name node to another host in accumulo 1.6
Posted by Josh Elser <jo...@gmail.com>.
Also, this list strips off images. You'll be better off sending a link to
some external hosting svc for image sharing.
On Jan 16, 2015 7:30 PM, "Calvin Feder" <ca...@argyledata.com> wrote:
> We need to move the HDFS name node of our cluster to a new host (the old
> host is experiencing hardware issues and may die at any time). We need to
> keep the current accumulo data. We are trying to do it in a test
> environment and running into numerous issues. We stopped the cluster, moved
> the name node to a new host, and restarted the cluster. We changed the
> entry in zookeeper for the accumulo root tablet to the new location, and
> got accumulo to start. We see all of the tables we expected to see in
> accumulo, and they show up in the monitor as “online”, but the monitor
> shows that no tablets or entries exist in those tables, and obviously we
> cannot scan them.
>
>
>
> This is an example of what we are seeing:
>
>
>
>
>
> Has anyone tried to do something similar? What are we missing?
>
>
>
> Thank you,
>
> *___________________________________________*
>
> *Calvin Feder *| *QA Lead | Argyle Data, Inc.*
>
> [image: cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7]
>
>
>
>
>
>
>
> ------------------------------
>
> This message is for the designated recipient only and may contain
> privileged, proprietary, or otherwise confidential information. If you have
> received it in error, please notify the sender immediately and delete the
> original. Any other use of the e-mail by you is prohibited. Thank you in
> advance for your cooperation.
> ------------------------------
>
RE: Moving HDFS name node to another host in accumulo 1.6
Posted by dl...@comcast.net.
In 1.6, accumulo.metadata entries are absolute URLs. You can change the existing URLs using the instance.volume.replacments[1] parameter.
[1] http://accumulo.apache.org/1.6/accumulo_user_manual.html#_instance_volumes_replacements
> -----Original Message-----
> From: webmaster@webmaster.ms [mailto:webmaster@webmaster.ms] On
> Behalf Of Denis
> Sent: Friday, January 16, 2015 7:53 PM
> To: user@accumulo.apache.org
> Subject: Re: Moving HDFS name node to another host in accumulo 1.6
>
> Hello
>
> Table "accumulo.metadata" contains absolute urls in a form of hdfs://old-
> name-node:8020/....
> I do not know how to work around it, except of patching accumulo code
> adding .replace(OLD_NODE, NEW_NODE) into relevant places
>
> On 1/17/15, Calvin Feder <ca...@argyledata.com> wrote:
> > We need to move the HDFS name node of our cluster to a new host (the
> > old host is experiencing hardware issues and may die at any time). We
> > need to keep the current accumulo data. We are trying to do it in a
> > test environment and running into numerous issues. We stopped the
> > cluster, moved the name node to a new host, and restarted the cluster.
> > We changed the entry in zookeeper for the accumulo root tablet to the
> > new location, and got accumulo to start. We see all of the tables we
> > expected to see in accumulo, and they show up in the monitor as
> > "online", but the monitor shows that no tablets or entries exist in
> > those tables, and obviously we cannot scan them.
> >
> > This is an example of what we are seeing:
> >
> > [cid:image002.png@01D031A9.978AE280]
> >
> > Has anyone tried to do something similar? What are we missing?
> >
> > Thank you,
> > ___________________________________________
> > Calvin Feder | QA Lead | Argyle Data, Inc.
> > [cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7]
> >
> >
> >
> >
> >
> > ________________________________
> >
> > This message is for the designated recipient only and may contain
> > privileged, proprietary, or otherwise confidential information. If you
> > have received it in error, please notify the sender immediately and
> > delete the original. Any other use of the e-mail by you is prohibited.
> > Thank you in advance for your cooperation.
> >
> > ________________________________
> >
Re: Moving HDFS name node to another host in accumulo 1.6
Posted by Denis <de...@camfex.cz>.
Hello
Table "accumulo.metadata" contains absolute urls in a form of
hdfs://old-name-node:8020/....
I do not know how to work around it, except of patching accumulo code
adding .replace(OLD_NODE, NEW_NODE) into relevant places
On 1/17/15, Calvin Feder <ca...@argyledata.com> wrote:
> We need to move the HDFS name node of our cluster to a new host (the old
> host is experiencing hardware issues and may die at any time). We need to
> keep the current accumulo data. We are trying to do it in a test
> environment and running into numerous issues. We stopped the cluster, moved
> the name node to a new host, and restarted the cluster. We changed the
> entry in zookeeper for the accumulo root tablet to the new location, and got
> accumulo to start. We see all of the tables we expected to see in accumulo,
> and they show up in the monitor as "online", but the monitor shows that no
> tablets or entries exist in those tables, and obviously we cannot scan
> them.
>
> This is an example of what we are seeing:
>
> [cid:image002.png@01D031A9.978AE280]
>
> Has anyone tried to do something similar? What are we missing?
>
> Thank you,
> ___________________________________________
> Calvin Feder | QA Lead | Argyle Data, Inc.
> [cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7]
>
>
>
>
>
> ________________________________
>
> This message is for the designated recipient only and may contain
> privileged, proprietary, or otherwise confidential information. If you have
> received it in error, please notify the sender immediately and delete the
> original. Any other use of the e-mail by you is prohibited. Thank you in
> advance for your cooperation.
>
> ________________________________
>
RE: Moving HDFS name node to another host in accumulo 1.6
Posted by dl...@comcast.net.
What Accumulo version?
What Hadoop version?
From: Calvin Feder [mailto:calvin.feder@argyledata.com]
Sent: Friday, January 16, 2015 7:29 PM
To: user@accumulo.apache.org
Subject: Moving HDFS name node to another host in accumulo 1.6
We need to move the HDFS name node of our cluster to a new host (the old
host is experiencing hardware issues and may die at any time). We need to
keep the current accumulo data. We are trying to do it in a test
environment and running into numerous issues. We stopped the cluster, moved
the name node to a new host, and restarted the cluster. We changed the
entry in zookeeper for the accumulo root tablet to the new location, and got
accumulo to start. We see all of the tables we expected to see in accumulo,
and they show up in the monitor as "online", but the monitor shows that no
tablets or entries exist in those tables, and obviously we cannot scan them.
This is an example of what we are seeing:
Has anyone tried to do something similar? What are we missing?
Thank you,
___________________________________________
Calvin Feder | QA Lead | Argyle Data, Inc.
cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7
_____
This message is for the designated recipient only and may contain
privileged, proprietary, or otherwise confidential information. If you have
received it in error, please notify the sender immediately and delete the
original. Any other use of the e-mail by you is prohibited. Thank you in
advance for your cooperation.
_____
Re: Moving HDFS name node to another host in accumulo 1.6
Posted by Sean Busbey <bu...@cloudera.com>.
Your best bet is take the opportunity to transition to HDFS-HA. That will
move all of your Accumulo operations under an HDFS nameservice, which will
then let you move namenodes in the future without more work.
We don't currently have good docs on how to handle this transition, but we
do have my working notes from the last time I did it. For info
see https://issues.apache.org/jira/browse/ACCUMULO-2793
Please let us know if you run into any trouble using those instructions in
your test environment.
On Fri, Jan 16, 2015 at 4:29 PM, Calvin Feder <ca...@argyledata.com>
wrote:
> We need to move the HDFS name node of our cluster to a new host (the old
> host is experiencing hardware issues and may die at any time). We need to
> keep the current accumulo data. We are trying to do it in a test
> environment and running into numerous issues. We stopped the cluster, moved
> the name node to a new host, and restarted the cluster. We changed the
> entry in zookeeper for the accumulo root tablet to the new location, and
> got accumulo to start. We see all of the tables we expected to see in
> accumulo, and they show up in the monitor as “online”, but the monitor
> shows that no tablets or entries exist in those tables, and obviously we
> cannot scan them.
>
>
>
> This is an example of what we are seeing:
>
>
>
>
>
> Has anyone tried to do something similar? What are we missing?
>
>
>
> Thank you,
>
> *___________________________________________*
>
> *Calvin Feder *| *QA Lead | Argyle Data, Inc.*
>
> [image: cid:A79ABAC4-829A-4B95-AFE2-C09282CD7AE7]
>
>
>
>
>
>
>
> ------------------------------
>
> This message is for the designated recipient only and may contain
> privileged, proprietary, or otherwise confidential information. If you have
> received it in error, please notify the sender immediately and delete the
> original. Any other use of the e-mail by you is prohibited. Thank you in
> advance for your cooperation.
> ------------------------------
>
--
Sean