Impala Load Balancing with Amazon Elastic Load Balancer

In a previous post, we explained how to configure a proxy server to provide load balancing for the Impala daemon. The proxy software used was HAproxy, a free, open source load balancer. This post will demonstrate how to use Amazon’s Elastic Load Balancer (ELB) to perform Impala load balancing when running in Amazon’s Elastic Compute Cloud (EC2).

Details

Similar to HAproxy, an Elastic Load Balancer is a reverse proxy that will take incoming TCP connections and distribute them amongst a set of EC2 instances. This is done partly for fault tolerance and partly for load distribution. Cloudera’s Using Impala through a Proxy for High Availability details how load balancing applies to part of Impala.

To summarize, the proxy will allow us to configure our Impala clients (Hue, Tableau, etc) with a single hostname and port. This well-known hostname will not have to be changed out if there were to be a failure of one or multiple Impala daemons. It will also distribute the load of running a query’s coordinator processing amongst the Impala daemons.

The steps we will take to set up the Impala load balancer are:

  1. Configure security groups.
  2. Create the ELB.
  3. Configure the ELB.
  4. Add instances to the ELB.
  5. Test the ELB.
  6. Configure Impala.

Requirements

The following examples require that you have the AWS CLI software installed and configured on your system. This can be on your Windows, Mac, or Linux workstation/laptop or on a Linux host running elsewhere. Spinning up an EC2 instance running Amazon Linux might be the fastest way to get the tools.

Of course, you will need an AWS account and IAM privileges to create both the security groups and the ELB.

Implementation

The following code is run in a shell (bash or cmd.exe) on the system with the AWS CLI tools.

First, we will need to name the cluster that is running the Impala service. This will be used to name the ELB and security groups. Then, we will create some variables that will hold the ID numbers of existing AWS infrastructure. Lastly, we use an additional variable named $OPTS for general AWS CLI options we may need.

We will have to look up the VPC ID, the ID of the subnet where we will be deploying the load balancer, and the instance IDs of the Hadoop cluster workers that are running the Impala Daemon. The ELB can reside in the same subnet as the Hadoop cluster or it can be placed in a separate subnet. It is advisable to keep the ELB in the same Availability Zone as the cluster. (You are deploying your Hadoop cluster instances to the same AZ, right?)

CLNAME=Cluster1
VPCID=""
SUBNETID=""
INSTANCES=""
#OPTS="--profile default --region us-west2"

If you have tagged your objects appropriately, you can use aws ec2 describe-* to look them up programmatically. These are examples. Your environment will be different.

# Return the VPCID of the VPC named "Cluster1".
VPCID=$(aws $OPTS ec2 describe-vpcs --output text --query 'Vpcs[*].VpcId' \
  --filter Name=tag:Name,Values="${CLNAME}")

# Return the SUBNETID of the subnet named "Cluster1 Private subnet 0".
SUBNETID=$(aws $OPTS ec2 describe-subnets --output text \
  --query 'Subnets[*].SubnetId' \
  --filter Name=tag:Name,Values="${CLNAME} Private subnet 0")

# Return the instance IDs of instances tagged "env=Cluster1" and "type=worker".
INSTANCES=$(aws $OPTS ec2 describe-instances --output text \
  --query 'Reservations[*].Instances[*].InstanceId' \
  --filters "Name=tag:env,Values=${CLNAME}" Name=tag:type,Values=worker)

Step 1

Second, we will define new security groups. The first group will allow client initiated traffic to reach the ELB. This should be locked down to something smaller than “everything” (0.0.0.0/0) especially if you have your cluster on the big, bad Internet.
The second group will allow the ELB to reach the Impala daemons running on the Hadoop cluster worker nodes.

echo "Allow Impala connections from clients to the load balancer."
FRONTEND=$(aws $OPTS ec2 create-security-group --output text --vpc-id $VPCID \
  --group-name "${CLNAME} Impala FE" --description "Impala Front-End Traffic")
aws $OPTS ec2 authorize-security-group-ingress --group-id $FRONTEND \
  --protocol tcp --port 21000 --cidr 0.0.0.0/0
aws $OPTS ec2 authorize-security-group-ingress --group-id $FRONTEND \
  --protocol tcp --port 21050 --cidr 0.0.0.0/0

echo "Allow Impala connections from the load balancer to the cluster."
BACKEND=$(aws $OPTS ec2 create-security-group --output text --vpc-id $VPCID \
  --group-name "${CLNAME} Impala BE" --description "Impala Back-End Traffic")
aws $OPTS ec2 authorize-security-group-ingress --group-id $BACKEND \
  --protocol tcp --port 21000 --source-group $FRONTEND
aws $OPTS ec2 authorize-security-group-ingress --group-id $BACKEND \
  --protocol tcp --port 21050 --source-group $FRONTEND

Then we will add each EC2 instance to the new $BACKEND security group.

for INSTANCEID in $INSTANCES; do
  GROUPID=$(aws $OPTS ec2 describe-instance-attribute --instance-id $INSTANCEID \
    --attribute groupSet --output text --query 'Groups[*].GroupId')
  aws $OPTS ec2 modify-instance-attribute --instance-id $INSTANCEID \
    --groups $GROUPID $BACKEND
done

Step 2

Next we get to the meat of this post: creating the load balancer. We will create an ELB with the name “elb-impala-Cluster1” and tell it to listen on ports 21000/TCP and 21050/TCP. The ELB will reside on subnet $SUBNETID and be a member of the security group $FRONTEND. This ELB will be internal/private and will not be available on the Internet. You can change this with the --scheme argument.

aws $OPTS elb create-load-balancer \
  --load-balancer-name elb-impala-${CLNAME} \
  --listeners \
  "Protocol=TCP,LoadBalancerPort=21000,InstanceProtocol=TCP,InstancePort=21000" \
  "Protocol=TCP,LoadBalancerPort=21050,InstanceProtocol=TCP,InstancePort=21050" \
  --subnets $SUBNETID \
  --security-groups $FRONTEND \
  --scheme internal

Step 3

After creation, we will modify some of the ELB configuration. This command will modify the Connection Idle Timeout value to 3600 seconds. It will also set up logging to go to a previously created S3 bucket named “Cluster1-logs” where files prefixed with “Cluster1-Impala” will be written every 60 minutes.

aws $OPTS elb modify-load-balancer-attributes \
  --load-balancer-name elb-impala-${CLNAME} \
  --load-balancer-attributes \
  AccessLog={Enabled=true,S3BucketName=${CLNAME}-logs,EmitInterval=60,S3BucketPrefix=${CLNAME}-Impala,ConnectionSettings={IdleTimeout=3600}

Further modifications to the ELB will add a health check for the ELB to determine if individual instances are available. The ELB will connect to port 21000/TCP every 30 seconds to test if the instance application is listening. The individual checks will timeout after 5 seconds with no response. The instance will be considered to have failed after two failed checks. The instance will return to a healthy status after five successful checks.

aws $OPTS elb configure-health-check \
  --load-balancer-name elb-impala-${CLNAME} \
  --health-check \
  Target=TCP:21000/png,Interval=30,Timeout=5,UnhealthyThreshold=2,HealthyThreshold=5

Step 4

Finally, we will attach the Hadoop worker instances to the ELB and load balancing will begin to be available.

aws $OPTS elb register-instances-with-load-balancer \
  --load-balancer-name elb-impala-${CLNAME} \
  --instances $INSTANCES

Lets not forget to look up the all-important DNS name that we will be using to talk to the ELB. We will use this to configure our client applications and for impala-shell.

ELBDNSNAME=$(aws $OPTS elb describe-load-balancers --load-balancer-names \
  elb-impala-${CLNAME} --output text --query 'LoadBalancerDescriptions[*].DNSName')
echo "*** SAVE ME ***"
echo "ELBDNSNAME : ${ELBDNSNAME}"
ELBDNSNAME=${ELBDNSNAME}

Testing

Step 5

We will test our implementation to confirm that it works to our expectation.

for (( i = 0 ; i < 10; i++ )); do
  impala-shell -i ${ELBDNSNAME} -q 'SELECT pid();' 2>&1 | grep Coordinator:
done

You should get output similar to the following which shows that we are connecting to a new coordinator each time:

Query submitted at: 2017-06-23 19:53:00 (Coordinator: http://ip-10-30-1-35.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:01 (Coordinator: http://ip-10-30-1-4.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:01 (Coordinator: http://ip-10-30-1-46.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:01 (Coordinator: http://ip-10-30-1-33.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:01 (Coordinator: http://ip-10-30-1-10.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:01 (Coordinator: http://ip-10-30-1-35.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:02 (Coordinator: http://ip-10-30-1-4.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:02 (Coordinator: http://ip-10-30-1-46.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:02 (Coordinator: http://ip-10-30-1-33.ec2.internal:25000)
Query submitted at: 2017-06-23 19:53:02 (Coordinator: http://ip-10-30-1-10.ec2.internal:25000)

Configure Impala

Step 6

Technically, there is nothing you need to configure in Impala. At least not on an insecure (non-Kerberized) cluster. You do need to tell other applications in your Hadoop distribution about the load balancer.

From Cloudera’s Using Impala through a Proxy for High Availability:

On systems managed by Cloudera Manager, on the page Impala > Configuration > Impala Daemon Default Group, specify a value for the Impala Daemons Load Balancer field. Specify the address of the load balancer in host:port format. This setting lets Cloudera Manager route all appropriate Impala-related operations through the proxy server.

Security

Since we at Clairvoyant tend to do a lot of security-enabled Hadoop deployments, it makes sense to describe how to get TLS enabled on the ELB.

Amazon has a service called AWS Certificate Manager. This service lets you provision a CA-signed TLS certificate onto the ELB with very little effort and will automatically update the certificate for you before it expires.

First we will request a certificate. Set the variable $DNAME to the name of the fully qualified domain name that you are using for the certificate. Then we will add tags so that we can provide a simple name.

DNAME=

ARN=$(aws $OPTS acm request-certificate --domain-name $DNAME \
  --subject-alternative-names $ELBDNSNAME --output text)
aws $OPTS acm add-tags-to-certificate --certificate-arn $ARN --tags \
  Name=tag:Name,Values="${CLNAME}"

At this point, the certificate is not yet issued. Emails have been sent to the domain contacts for approval of the request. Once the request is approved, we can continue with assigning the certificate to the ELB. We can list the certificates and watch to see if it has been approved. If there is any output to this command, it means the request has been approved by the domain owner.

aws $OPTS acm list-certificates --certificate-statuses ISSUED --output text \
  | grep $ARN

Lastly, we will modify the ELB from TCP mode to SSL mode and tell it to use the TLS certificate on ports 21000/TCP and 21050/TCP.

aws $OPTS delete-load-balancer-listeners \
  --load-balancer-name elb-impala-${CLNAME} --load-balancer-ports 21000
aws $OPTS elb create-load-balancer-listeners \
  --load-balancer-name elb-impala-${CLNAME} \
  --listeners \
  "Protocol=SSL,LoadBalancerPort=21000,InstanceProtocol=SSL,InstancePort=21000,SSLCertificateId=${ARN}"

aws $OPTS delete-load-balancer-listeners \
  --load-balancer-name elb-impala-${CLNAME} --load-balancer-ports 21050
aws $OPTS elb create-load-balancer-listeners \
  --load-balancer-name elb-impala-${CLNAME} \
  --listeners \
  "Protocol=SSL,LoadBalancerPort=21050,InstanceProtocol=SSL,InstancePort=21050,SSLCertificateId=${ARN}"

You should now have an Amazon-signed TLS certificate protecting your ELB traffic. To confirm, run the following command and look for something like issuer= /C=US/O=Amazon/OU=Server CA 1B/CN=Amazon.

openssl s_client -connect ${ELBDNSNAME}:21000 -nbio /dev/null \
  | openssl x509 -noout -issuer

Thats it. Happy load balancing!

Impala High Availability

Impala daemon is a core component in the Impala architecture. The daemon process runs on each data node  and is the process to which the clients (Hue,JDBC,ODBC) connect to issue queries.  When a query gets submitted to an Impala daemon ,  that node serves as the coordinator node for that query. Impala daemon acting as the co-ordinator parallelizes the queries and distributes work to other nodes in the Impala cluster. The other nodes transmit partial results back to the coordinator, which constructs the final result set for a query.

It is a recommended practice to run Impalad on each of the Data nodes in a cluster , as Impala takes advantage of the data locality while processing its queries. So most of the time the Impala clients connect to any of the data nodes to run their queries. This might create a single point of failure for the clients if the clients are always issuing queries to a single data node. In addition to that the node acting as a coordinator node for each Impala query potentially requires more memory and CPU cycles than the other nodes that process the query. For clusters running production workloads,  High Availability from the Impala clients standpoint and load distribution across the nodes can be achieved by  having a proxy server or load-balancer to issue queries to impala daemons using a round-robin scheduling.

HAProxy  is free, open source load balancer that can be used as a proxy-server or load balancer to distribute the load across different impala daemons. The high level architecture for this setup looks like below.

impala-high-availability

 

Install the load balancer:

HAProxy can installed and configured on Red Hat Enterprise Linux system and Centos OS using the following instructions.

yum install haproxy

Set up the configuration file: /etc/haproxy/haproxy.cfg.

See the following section for a sample configuration file

global
log         127.0.0.1 local2

chroot      /var/lib/haproxy
pidfile     /var/run/haproxy.pid
maxconn     4000
user       haproxy
group      haproxy
daemon

# turn on stats unix socket
stats socket /var/lib/haproxy/stats

#---------------------------------------------------------------------
# common defaults that all the 'listen' and 'backend' sections will
# use if not designated in their block
#---------------------------------------------------------------------
defaults
mode                   tcp
log                    global
 
retries               3
timeout connect       50000s
timeout client        50000s
timeout server        50000s
maxconn               3000

#---------------------------------------------------------------------
# main frontend which proxys to the backends - change the port
# if you want
#---------------------------------------------------------------------
frontend main *:5000
acl url_static       path_beg       -i /static /images /javascript/stylesheets
acl url_static       path_end       -i .jpg .gif .png .css .js

use_backend static          if url_static
default_backend            impala

#---------------------------------------------------------------------
#static backend for serving up images, stylesheets and such
#---------------------------------------------------------------------
backend static
    balance     roundrobin
    server      static 127.0.0.1:4331 check

#---------------------------------------------------------------------
#round robin balancing between the various backends
#---------------------------------------------------------------------
backend impala
    mode          tcp
    option        tcplog
    balance       roundrobin
    #balance      leastconn
	#---------------------------------------------------------------------
	# Replace the ip addresses with your client nodes ip addresses
	#---------------------------------------------------------------------
    server client1 192.168.3.163:21000
    server client2 192.168.3.164:21000
    server client3 192.168.3.165:21000

Run the following command after done the changes

service haproxy reload;

Note:

The key configuration options are balance and server in the backend impala section. As well as the timeout configuration options in the defaults section. The server with the lowest number of connections receives the connection only when the balance parameter is set to leastconn. If balance parameter is set to roundrobin, the proxy server can issue queries to each connection uses a different coordinator node.

  1. On systems managed by Cloudera Manager, on the page Impala > Configuration > Impala Daemon Default Group, specify a value for the Impala Daemons Load Balancer field. Specify the address of the load balancer in host:port format. This setting lets Cloudera Manager route all appropriate Impala-related operations through the proxy server.
  2. For any scripts, jobs, or configuration settings for applications that formerly connected to a specific datanode to run Impala SQL statements, change the connection information (such as the -i option inimpala-shell) to point to the load balancer instead.

Test Impala through a Proxy for High Availability:

Manual testing with HAProxy:

Stop the impala daemon service one by one and run the queries, check the impala high availability is working fine or not.

Test the impala high availability with shell script using HAProxy:

Run the following shell script and test the impala high availability using HAProxy.

Note: Please change the ‘table_name’ and ‘database_name’ placeholders.

for (( i = 0 ; i < 5; i++ ))
do
    impala-shell -i localhost:5000 -q "select * from {table_name}" -d {database_name}
done

Result: Run the above script and find the usage of load balancer.

  • Query should be executing on different impala daemon nodes for each iteration (when balance is roundrobin).