We have a Cloudera cluster up and running with an h2o instance although it appears to be running off h2o.jar (which as I understand it--please correct me if incorrect) is the stand-alone h2o. I can connect, but it will not load any files from our HDFS. (all of this i can see via 'ps' on edge node.
So I started an instance with h2odriver.jar
java -jar /path/to/h2odriver.jar -nodes 2 -mapperXmx 5g -output /my/hdfs/dir
I get several output/callback addresses:
[Possible callback IP address: 10.96.243.46:33728]
[Possible callback IP address: 127.0.0.1]
Using mapper->driver callback IP address and port: 10.96.243.46:33728
So I fire up python and try and connect (same thing happens if I use 10.96.243.46):
>>>h2o.connection(ip='127.0.0.1', port='33728')
and get
'Connecting to H2O server at http://127.0.0.1:33728..... failed.
H2OConnectionError: COuld not estalich link to the H2O cloud http://127.0.0.1:33728 after 5 retries
...
Failed to establish a new connection:[Errno 111] Connection refused',))`
Thing is on my screen with the H2O jar/java job I can see:
`MapperToDriverMessage: Read invalid type (G) from socket, ignoring...
MapperToDriverMessage: read: Unknown Type `
I cannot figure out how to launch h2o in cluster mode and have it access our hdfs system or even connect. I can connect to the h2o.jar version, but that sees no hdfs (it can see the filesystem of the edgenode). What is the proper way to launch H2O so that it can see the attached HDFS system (We are running Cloudera 5.7 in a enterprise environment, Python is 3.6, H2O is 3.10.0.6 and I know we have a ton of firewalls/security-- i beleive we are setup through LDAP
You are correct that h2o.jar is meant to be the standalone version of H2O which is not meant for connecting to HDFS.
Using the appropriate h2odriver.jar for your particular hadoop distribution is the way to go.
The correct beginner instructions can be found here:
go to http://www.h2o.ai/download/
choose H2O "Latest Stable Release"
choose tab "Install on Hadoop"
It says to run the following command:
hadoop jar h2odriver.jar -nodes 1 -mapperXmx 6g -output hdfsOutputDirName
[ Note this is "hadoop jar", not "java -jar" as written in the question. ]
You should see output like this:
Determining driver host interface for mapper->driver callback...
[Possible callback IP address: 172.16.2.181]
[Possible callback IP address: 127.0.0.1]
...
Waiting for H2O cluster to come up...
H2O node 172.16.2.188:54321 requested flatfile
Sending flatfiles to nodes...
[Sending flatfile to node 172.16.2.188:54321]
H2O node 172.16.2.188:54321 reports H2O cluster size 1
H2O cluster (1 nodes) is up
(Note: Use the -disown option to exit the driver after cluster formation)
Open H2O Flow in your web browser: http://172.16.2.188:54321
(Press Ctrl-C to kill the cluster)
Blocking until the H2O cluster shuts down...
Then point your web browser to the place where it says to "Open H2O Flow in your web browser".
(The other addresses in the output are diagnostics, and not for end users.)
In this case, the python connection command would be:
h2o.connect(ip = '172.16.2.188', port = 54321)
I recommend going to Flow in a web browser, start importing a file by typing in "hdfs://", and seeing if autocompletion works. If it does, your HDFS connection is working.
Related
I've setup opscenter on one of cassandra cluster nodes. After installation, when setting up my cluster, I tried installation of datastax agent on all the cluster nodes via UI, but it failed. So, I had to install the agents manually.
After manually installing the agents, the node in which opscenter is installed is able to connect, but not the other nodes. It still says, "2 agents failed to connect". What could be the issue?
PS : My cassandra cluster is setup on AWS in ubuntu
My agent.log file looks like this
ERROR [os-metrics-9] 2015-07-27 07:04:43,390 Long os-stats collector failed: Cannot run program "iostat": error=2, No such file or directory
ERROR [os-metrics-7] 2015-07-27 07:04:43,391 Long os-stats collector failed: Cannot run program "iostat": error=2, No such file or directory
ERROR [os-metrics-8] 2015-07-27 07:04:53,391 Long os-stats collector failed: Cannot run program "iostat": error=2, No such file or directory
ERROR [os-metrics-3] 2015-07-27 07:04:53,392 Long os-stats collector failed: Cannot run program "iostat": error=2, No such file or directory
ERROR [StompConnection receiver] 2015-07-27 07:05:02,946 failed connecting to **.**.**.**:61620:java.net.ConnectException: Connection timed out
You have to set the stomp_interface in the address.yaml like
stomp_interface: <ip-address>
After agent restart it should be connected.
As your agent have been able to connect from the same box where opscenter is installed, so it sounds like :
You might have not configured your firewall properly. If you please try by disabling firewall on all your boxes.
You may have multiple interfaces and C* installation picked up an undesired interface. So run ifconfig or ip command on all of your instances and check with C* yaml.
About iostat failure message : You have not install sysstat pkg. Seems, you have not install dependencies as part of DSE install.
The agents uses iostat to collect some information from disks. If it cant find it you will get that error but it just means those metrics will be missing some os metrics (likely a lot of disk and cpu metrics will be missing)
These are some useful configurations that you should keep in mind when starting the agent manually in the conf/address.yaml file:
###A name for the node to use as a label throughout OpsCenter.
alias:
###Reachable IP address of the opscenterd machine. The connection made will be on stomp_port. Internal IP in this case
stomp_interface:
###Port for the agent's HTTP service (default: 61621).
#api_port: 61621
###The stomp_port used by opscenterd. == Must match with the 'incoming_port' in opscenter.conf
stomp_port: 61620
###The IP used to identify the node.
local_interface: 100.73.158.44
###The IP that the agent HTTP server listens on.
agent_rpc_interface:
###Host used to connect to local JMX server.
jmx_host: 100.73.158.44
###Whether or not to use SSL communication between the agent and opscenterd.
use_ssl: 1
To solve the "Cannot run program 'iostat'" error, do this:
sudo apt-get install sysstat
I'm trying to install the enterprise edition of neo4j on an existing EC2 (Amazon linux) instance. So far I've
wget "link to enterprise"
untar the file
renamed and moved the folder to NEO4J_HOME
then went into the config files for neo4j.properties to make the following changes:
# Enable shell server so that remote clients can connect via Neo4j shell.
remote_shell_enabled=true
# The network interface IP the shell will listen on (use 0.0.0 for all interfaces)
remote_shell_host=127.0.0.1
# The port the shell will listen on, default is 1337
remote_shell_port=1337
EDITED Christophe Willemsen pointed out that for my original error, I had forgotten to restart the server at that point but I was still unable to access the web server while it was running. So to make it more clear, I've edited the remaining post:
I went to neo4j-server.properties and uncommented:
org.neo4j.server.webserver.address=0.0.0.0
And start the server
NEO4J_HOME/bin/neo4j start
WARNING: Max 1024 open files allowed, minimum of 40 000 recommended. See the Neo4j manual.
Using additional JVM arguments: -server -XX:+DisableExplicitGC -Dorg.neo4j.server.properties=conf/neo4j-server.properties -Djava.util.logging.config.file=conf/logging.properties -Dlog4j.configuration=file:conf/log4j.properties -XX:+UseConcMarkSweepGC -XX:+CMSClassUnloadingEnabled -XX:-OmitStackTraceInFastThrow
Starting Neo4j Server...WARNING: not changing user
process [28557]... waiting for server to be ready..... OK.
http://localhost:7474/ is ready.
checking the status:
NEO4J_HOME/bin/neo4j status
Neo4j Server is running at pid 28557
I can run the shell but the when I go to localhost 7474 I still can not connect
Any help would be appreciative. The only tutorial or help I've found assumed I was starting from scratch with a new instance. If someone could provide some instructions for installing or fix my configuration that would be great.
Thanks!
You have to edit neo4j-server.properties and uncomment the line with:
org.neo4j.server.webserver.address=0.0.0.0
So that the db listens on an external interface not just localhost, and you have to open the port (7474) in your firewall rules.
Make sure to secure access to the db though:
http://neo4j.com/docs/stable/security-server.html
While debugging I realised that confd doesn't pick up the keys and my journal looks like this:
Sep 18 18:31:50 ip-10-171-54-76.ec2.internal docker[24891]: [nginx] waiting for confd to refresh nginx.conf
Sep 18 18:31:56 ip-10-171-54-76.ec2.internal docker[24891]: 2014-09-18T18:31:56Z 9122c7a54edc confd[9572]: ERROR 501: All the given peers are not reachable (Tried to connect to each peer twice and failed) [0]
I use nsenter to log in to the running container to run some experiments for debugging purposes. I ran this command
confd -onetime -node 172.17.42.1:4001 -config-file /etc/confd/conf.d/nginx.toml
Then received this error as above
confd[12894]: ERROR 501: All the given peers are not reachable (Tried to connect to each peer twice and failed) [0]
I am totally clueless at this point. I am using EC2 with the stable version of CoreOS and I am sure that etcd is running on the host. Also, I can ping the host from inside the container successfully.
Any ideas on what's wrong?
Assistance will be much appreciated.
This error indicates that your etcd cluster isn't operating correctly, so confd has nothing to watch. It has probably lost quorum. The logs (journalctl -u etcd) should indicate what happened.
I have to do distributed testing using JMeter. The objective is to have multiple remote servers in AWS controlled by one local server send a file download request to another server in AWS.
How can I set up the different servers in AWS?
How can I connect to them remotely?
Can someone provide some step by step instructions on how to do it?
I have tried several things but keep running into connectivity issues across networks.
We had a similar task and we ran into a bunch of issues as well. Here are the details of the whole process and what we did to resolve the issues we encountered. Hope it helps.
We needed to send requests from 5 servers located in various regions of the world. So we launched 5 micro instances in AWS, each in a different region. We chose the regions to be as geographically apart as possible.
Remote (server) JMeters config
Here is how we set up each instance.
Installed java:
$ sudo apt-get update
$ sudo apt-get install default-jre
Installed JMeter:
$ mkdir jmeter
$ cd jmeter;
$ wget ftp://apache.mirrors.pair.com//jmeter/binaries/apache-jmeter-2.9.tgz
$ gunzip apache-jmeter-2.9.tgz;tar xvf apache-jmeter-2.9.tar
Edited the jmeter.properties file in the /bin folder of the JMeter installation and uncomment the line containing the server.rmi.localport setting. We changed the port to 50000.
server.rmi.localport=50000
Started JMeter server. Make sure the address and the port the server reports listening to are correct.
$ cd ~/jmeter/apache-jmeter-2.9/bin
$ vi jmeter-server
Local (client) JMeter config
Then we set up JMeter to run tests remotely on these instances on our local client machine:
Ensured to use the same version of JMeter as was running on the servers. Installed Java and JMeter as described above.
Enabled remote testing by editing the jmeter.properties file that can be found in the bin folder of the JMeter installation. The parameter remote_hosts needed to be set with the public DNS of the remote servers we were connecting to.
remote_hosts=54.x.x.x,54.x.x.x,54.x.x.x,54.x.x.x,54.x.x.x
We were now able to tell our client JMeter instance to run tests on any or all of our specified remote servers.
Issues and resolutions
Here are the issues we encountered and how we resolved them:
The client failed with:
ERROR - jmeter.engine.ClientJMeterEngine: java.rmi.ConnectException: Connection - refused to host: 127.0.0.1
It was due to the server host returning the private IP address as its address because of Amazon NAT.
We fixed this by setting the parameter RMI_HOST_DEF that the /usr/local/jmeter/bin/jmeter-server script includes in starting the server:
RMI_HOST_DEF=-Djava.rmi.server.hostname=54.xx.xx.xx
Now, the AWS instance returned the server’s external IP, and we could start the test.
When the server node attempted to return the result and tried to connect to the client, the server tried to connect to the external IP address of my local machine. But it threw a connection refused error:
2013/05/16 12:23:37 ERROR - jmeter.samplers.RemoteListenerWrapper: testStarted(host) java.rmi.ConnectException: Connection refused to host: xxx.xxx.xxx.xx;
We resolved this issue by setting up reverse tunnels at the client side.
First, we edited the jmeter.properties file in the /bin folder of the JMeter installation and uncommented the line containing the client.rmi.localport setting. We changed the port to 60000:
client.rmi.localport=60000
Then we connected to each of the servers using SSH, and setup a reverse tunnel to port 60000 on the client.
$ ssh -i ~/.ssh/54-x-x-x.us-east.pem -R 60000:localhost:60000 ubuntu#54.x.x.x
We kept each of these sessions open, as the JMeter server needs to be able to deliver the test results to the client.
Then we set up the JVM_ARGS environment variable on the client, in the jmeter.sh file in the /bin folder:
export JVM_ARGS="-Djava.rmi.server.hostname=localhost"
By doing this, JMeter will tell the servers to connect to localhost:60000 for delivering their results. This ends up being tunneled back to the client.
The SSH connections to the servers kept dropping after staying idle for a little bit. To prevent that from happening, we added a parameter to each of the SSH tunnel set up directing the client to wait 60 seconds before sending a null packet to the server to keep the connection alive:
$ ssh -i ~/.ssh/54-x-x-x.us-east.pem -o ServerAliveInterval=60 -R 60000:localhost:60000 ubuntu#54.x.x.x
(.ssh/config version of all required SSH settings:
Host 54.x.x.x
HostName 54.x.x.x
Port 22
User ubuntu
ServerAliveInterval 60
RemoteForward 127.0.0.1:60000 127.0.0.1:60000
IdentityFile ~/.ssh/54-x-x-x.us-east.pem
IdentitiesOnly yes
Just use ssh 54.x.x.x after setting this up.
)
I just went though this on openstack and found the same issues... no idea why the jmeter remoting documentation only covers half the required steps. You can do it without tunnels or touching the properties files.
You need
All nodes to advertise their public IP - on AWS/OS this defaults to the private IP
Ingress rules for the RMI port which defaults to 1099 - I use this
Ingress rules for the RMI "local" port which defaults to dynamic. Below I use 4001 for the client and 4000 for servers. The port can be the same but note the properties are different.
If you are using your workstation as the client you probably still need tunnels. Above Archana Aggarwal has good tips for tunnels.
Remote servers
Set java.rmi.server.hostname and server.rmi.localport inline or in the properties file.
jmeter-server -Djava.rmi.server.hostname=publicip -Dserver.rmi.localport=4000
Sneaky server on client
You can also run one on the same machine as the client. For clarity I've set java.rmi.server.hostname but left server.rmi.localport as dynamic
jmeter-server -Djava.rmi.server.hostname=localip
Client
Set java.rmi.server.hostname and client.rmi.localport inline or in the properties file. Use -R etc like so:
jmeter -n -t Test.jmx -Rremotepublicip1,remotepublicip2 -Djava.rmi.server.hostname=clientpublicip -Dclient.rmi.localport=4001 -GmypropA=1 -GmypropB=2 -lresults.jtl
When you go for distributed testing using JMeter in AWS, I would suggest you to use docker - which will help us with jmeter test infrastructure very quickly. This way we can also ensure that same version of java and jmeter are installed in all the instances of amazon which is very important of JMeter distributed testing.
Ensure that - you set below properties and ports are open for jmeter-server. [they do not have to be 1099,50000 exactly]
server.rmi.localport=50000
server_port=1099
java.rmi.server.hostname=SERVER_IP
for client
client.rmi.localport=60000
java.rmi.server.hostname=SERVER_IP - this step is very important as the container in aws instance will have their own IP address in the docker network - so master and slave can not communicate. So we explicitly set this property
More info:
http://www.testautomationguru.com/jmeter-distributed-load-testing-using-docker-in-aws/
WHAT AM I TRYING TO DO
Trying to setup a VCAP on a UBUNTU SERVER VM on my machine by following the steps mentioned at https://github.com/cloudfoundry/vcap/
WHAT IS THE ISSUE
Things seemed to be working fine but at step5 (https://github.com/cloudfoundry/vcap/#step-5-validate-that-you-can-connect-and-tests-pass) I got an exception while trying to execute the following command - vmc target api.vcap.me
The exception that I see on my console is:
Host is not available or is not valid: 'http://api.vcap.me'
Would you like see the response? [yN]: y
HTTP exception: Errno::ECONNREFUSED:No connection could be made because the target machine actively refused it. - connect(2)
ANY OTHER RELEVANT INFO
For some earlier experiments I was using MicroCloud (provided as a download by CloudFoundry). I am having issues in pointing my VMC to this Microcloud as well.
On the Micro Cloud console I see the following message:
To access your Micro Cloud Foundry instance, use:
vmc target http://api.agoel.cloudfoundry.me
When I run this vmc command from the Ruby Command Prompt setup on my Windows7 I get following error:
Host is not available or is not valid: 'http://api.agoel.cloudfoundry.me'
Would you like see the response? [yN]: y
HTTP exception: Errno::ETIMEDOUT:A connection attempt failed because the connected party did not properly respond after a period of time, or
ost has failed to respond. - connect(2)
WHATS DOES VMC INFO DISPLAY
I ran vmc info command on command prompt. It displayed following info
VMware's Cloud Application Platform
For support visit support DOT cloudfoundry DOT com
Target: http:// api DOT cloudfoundry DOT com (v0.999)
Client: v0.3.18
User: ankitgoel1987#gmail.com
Usage: Memory (1.1G of 2.0G total)
Services (2 of 16 total)
Apps (2 of 20 total)
MY SETUP DETAILS
Windows7 running on 4GB RAM
Microcloud from Cloudfoundry already installed (this was done as part of some other exercise. My recent experiment requires me to setup a Ubuntu server with VCAP on it. So this MicroCloud should not really matter)
vmc 0.3.18 (installed on my Windows7 machine)
ruby 1.9.2p290 (2011-07-09) [i386-mingw32]
add in your hosts files the following entry:
IP_of_ubuntu_server vcap.me api.vcap.me
If you want to avoid having to edit your hosts file every time you deploy a new app and depending on what virtualisation platform you are using you may be able to forward all traffic on port 80 for your own computer on to the VM.
*.vcap.me is set to resolve to 127.0.0.1 so this is an ideal solution. To do this you should set the network settings to NAT rather than Bridged (maybe you have done this already) and then set port 80 to forward to the IP of the guest OS. In VMWare Fusion for example this is as simple as editing a settings file.