Hi Marcel,
First thing to do is to check the logs on the tservers for any red
flags just prior to them going down - that might point out the
problem.
Failing that, I would recommend checking your memory settings. In
particular, you might need to bump up the memory allocated to each
tserver. Also ensure that your machines have enough memory to
satisfy all the various processes running on them.
You might also get some additional configuration tips on the
accumulo mailing list.
Thanks,
Emilio
On 07/01/2015 11:18 AM, Marcel wrote:
Thanks, Quickstart is working.
Now I´m struggling with the ingestion of GDELT dataset into
accumulo. I want to ingest the whole GDELT dataset (approximately
100 GB). When trying this, my tabletservers were killed after a
while. So I started with one file and the MapReduce job finished
successfully. But the second MapReduce job (for another day)
failed, because the tabletservers were down again. It seems that
the tabletserver are not stable. Following errors where thrown
corresponding to the log file:
org.apache.thrift.transport.TTransportException:
java.io.IOException: Connection reset by peer
unable to get tablet server status
node2-scads05:10011[14e49e8888b0001]
org.apache.thrift.transport.TTransportException:
java.net.ConnectException: Connection refused
Do you have any clue, how to handle this problem?
Thanks again.
Marcel Jacob.
Am 30.06.2015 15:56, schrieb Jim
Hughes:
Hi Marcel,
Good news on a few fronts: First, the error is due to a
mismatch in the old Accumulo 1.6 branch and the GeoMesa
Quickstart code. Assuming that you are building the QuickStart
manually, you can edit QuickStart.java and change this line
(likely line 126) from
"Who:String:index=full",
to
"Who:String",
To be safe, you'll need to delete the data in the table you
specified on the command line when running the QuickStart jar.
You can do this by dropping that table or manually deleting the
entries in the Accumulo shell (with deletemany). Alternatively,
you can use a different table name for your next test.
The additional good news is that updating GeoMesa's Accumulo 1.6
support is on my short list of tasks for this week. We've made
a number of improvements since 1.0.0-rc3 and now.
By 'bounce Accumulo', Emilio meant that one may need to restart
Accumulo when the jars in lib/ext are changed. If you upgrade
later, you may need to do that and/or reingest data.
Thanks,
Jim
On 06/30/2015 09:36 AM, Marcel
wrote:
Hello,
I´ve got some problems with the Quick-Start example using
Accumulo 1.6 on CDH 5.4.2. The exception is very similiar like
the one here https://locationtech.org/mhonarc/lists/geomesa-users/msg00429.html
It also says:
java.lang.IllegalArgumentException: For input string: "full"
but I´m confused what is meant by "You may need to bounce accumulo for jar
changes to take effect."?
Both
geomesa-distributed-runtime.accumulo1.6.-1.0.0-rc.3-SNAPSHOT.jar
and joda-time-2.3.jar are located in lib/ext on my tablet
servers. I also restarted all services on my cluster.
Thanks in advance.
Best regards,
Marcel Jacob.
_______________________________________________
geomesa-users mailing list
geomesa-users@xxxxxxxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
http://www.locationtech.org/mailman/listinfo/geomesa-users
_______________________________________________
geomesa-users mailing list
geomesa-users@xxxxxxxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
http://www.locationtech.org/mailman/listinfo/geomesa-users
_______________________________________________
geomesa-users mailing list
geomesa-users@xxxxxxxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
http://www.locationtech.org/mailman/listinfo/geomesa-users
|
