Load Balancing

Load balancing refers to distributing incoming traffic across a group of backend server. In order to serve a high volume of concurrent requests from users or clients it is required to have a “traffic distributer” sitting in front of the backend servers and routing requests across all workers capable of fulfilling those requests.

The load balance is responsible for the following functions:

  • Distributes client requests or network load efficiently across multiple backend servers

  • Ensures high availability and reliability by sending requests only to servers that are online and can handle the request in a timely manner

  • Provides the flexibility to scale-up or scale-down as demand dictates

In the context of the Object Storage, a Proxy Virtual Controller is referred as a backend server.

How load is balanced in the Object Storage?

The NextGen Object Storage load-balancer has two modes of operation:

Internal load balancer (default)

Any NextGen Object Storage is provisioned with a built-in Load Balancer (referred as Internal Load Balancer) solution which will work seamlessly out of the box. The internal load balancer is fully configured to provide a secured, reliable and highly-available endpoint. Upon provisioning a new Object Storage instance, the system will be have a globally unique FQDN matching the Object Storage front-end network, along with a matching well-known CA certificate that will be used for a secured communication over HTTPS.

In case a public IP/additional VNI (Virtual Network Interface) is allocated to the object storage, the internal load balancer will be used to distribute the workload. The main difference between Frontend network and Public/VNI handling is the load balancer mode of operation:

  • Frontend - DRS (Direct Server Return), packets from the Object Storage Virtual Controller bypass the Load Balancer, maximizing egress throughput.

  • Public/VNI - The load balancer will be used as a gateway for all traffic from/to the Object Storage Virtual Controller.

internal-load-balancer

Load balancing algorithm

The load balancer is configured to use weighted least connection (wlc) - new connections go to the worker with the least number of connections. Reminder - Controller VCs & Storage VCs are excluded from handling client/users operations.

High availability

As all other core services in the system, the internal load balancer is highly available. The internal load-balancer service will be hosted in a Controller VC (vc-1) and in case of a failure will failover to vc-0 to ensure service continuity.

Load balancer metering

While the actual object storage operations will be distributed to the “backend servers”, in some cases the system administrator would like to review the amount of concurrent sessions their object storage is handling.

In the NextGen Object Storage the active session count can be reviewed in the management interface:

  1. Virtual Controllers view - navigate to the Virtual Controller view and select the virtual controller with vc-1 as ID. In the south pane, select the Frontend Metering tab.

  2. Performance section - under the system section, navigate to the Performance section. Expand the Virtual Controller > Controller > vc-1 view and drag and drop the Active Connections pane.

The Active Connections connections graph will provide a breakdown of connections within the object storage (per VC). The user can toggle between the graph view and a the table view using the TABLES view.

Using an external load balancer

Zadara’s Object Storage provides an easy way to integrate with an existing or newly provisioned software/hardware load balancer.

The instructions below are example for setting up an external load balancer to terminate SSL connections and distribute the across all VCs.

There are many load balancer solutions in the market, setting them all up is quite similar procedure. This appendix gives an example of HAproxy, an open-source TCP/HTTP load-balancing proxy server that can be found in www.haproxy.org

The external load-balancer recommended configuration below will allow the following:

  • SSL Termination is done on the external load balancer for both object operation API’s and GUI connections. Authentication connections are always handled in the Object Storage itself.

  • Custom TLS certificate located on the load balancer is used for TLS connections

  • Object operation connections are redirected to NextGen Object Storage proxy VC’s

  • Object operation connections are distributed between VC’s unevenly (proxy VCs to take more load than storage VC’s, and HA VCs to take the lowest load)

  • Redirected object operation connections will include the original client IP in a special header added by the load balancer (for logging in NextGen Object Storage proxy)

  • HTTP-based health check is performed by the load balancer to probe all NextGen Object Storage proxy VC’s

  • Authentication connections are redirected to the Object Storage floating IP (SSL pass-through terminated on the NextGen Object Storage, Custom TLS certificate must be uploaded to the NextGen Object Storage as well).

  • UI requests are redirected to the NextGen Object Storage floating IP (over port 8443)

  • Graphical statistics interface is enabled on the load balancer

Apply the following configuration to your NextGen Object Storage Settings:

  1. Set the internet-facing domain-name/IP of the external LoadBalancer as NextGen Object Storage API Hostname / IP (zadara-qa.com which resolves to the external LB IP 180.80.2.217, is set  in this example as NextGen Object Storage API Hostname)

  2. Upload your custom SSL certificate (will be used for authentication connections). The certificate should match the custom domain name.

  3. Set SSL Termination to “External”

HAProxy Installation and configuration instructions:

  • Install HA Proxy:

    sudo add-apt-repository -y ppa:vbernat/haproxy-1.5
    sudo apt-get update
    sudo apt-get install -y haproxy
    
  • Upload your custom SSL certificate to HAProxy server. In this example the certificate PEM file is placed under /etc/ssl/private/zadara\_custom.pem

  • Edit /etc/haproxy/haproxy.cfg to include the following:

    global
        maxconn 2048
        log /dev/log  local0
        log /dev/log  local1 notice
        chroot /var/lib/haproxy
        stats socket /run/haproxy/admin.sock mode 660 level admin
        stats timeout 30s
        user haproxy
        group haproxy
        daemon
        tune.ssl.default-dh-param 2048
        # Default SSL material locations
        ca-base /etc/ssl/certs
        crt-base /etc/ssl/private
        # Default ciphers to use on SSL-enabled listening sockets.
        # For more information, see ciphers(1SSL). This list is from:
        #  https://hynek.me/articles/hardening-your-web-servers-ssl-ciphers/ssl-default-bind-ciphers
        ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+3DES:!aNULL:!MD5:!DSS
        ssl-default-bind-options no-sslv3
    
     defaults
        log     global
        mode    http
        option  httplog
        option  dontlognull
        timeout connect 5000
        timeout client  50000
        timeout server  50000
        errorfile 400 /etc/haproxy/errors/400.http
        errorfile 403 /etc/haproxy/errors/403.http
        errorfile 408 /etc/haproxy/errors/408.http
        errorfile 500 /etc/haproxy/errors/500.http
        errorfile 502 /etc/haproxy/errors/502.http
        errorfile 503 /etc/haproxy/errors/503.http
        errorfile 504 /etc/haproxy/errors/504.http
        frontend fe-object-operations
        bind 180.80.2.217:443 ssl crt /etc/ssl/private/zadara\_custom.pem
        mode http
        default\_backend be-zios-object-operations
    
     frontend fe-auth
        bind 180.80.2.217:5000
        option tcplog
        mode tcp
        default\_backend be-floating-zios-auth
    
     frontend fe-gui
        bind 180.80.2.217:8443 ssl crt /etc/ssl/private/zadara\_custom.pem
        mode http
        default\_backend be-floating-zios-gui
    
     backend be-zios-object-operations
        mode http
        balance roundrobin
        option forwardfor
        option  httpclose
        option  httpchk HEAD /healthcheck HTTP/1.0
        server ziosStorageProxy0 190.90.2.102:8080 weight 10 check
        server ziosStorageProxy1 190.90.2.104:8080 weight 10 check
        server ziosStorageProxy2 190.90.2.114:8080 weight 50 check
        server ziosProxyOnly3 190.90.2.106:8080 weight 100 check
        server ziosProxyOnly4 190.90.2.109:8080 weight 100 check
    
     backend be-floating-zios-auth
        mode tcp
        server ziosFloating 190.90.2.118:5000
    
     backend be-floating-zios-gui
        mode http
        server ziosFloating 190.90.2.118:80
    
     listen stats \*:1936
        stats enable
        stats uri /
        stats auth zadara:zadara
    
  • Enable HAProxy logging (Optional)

    • Edit rsyslog.conf

      sudo vi /etc/rsyslog.conf
      # provides UDP syslog reception
      $ModLoad imudp
      $UDPServerRun 514
      # provides TCP syslog reception
      $ModLoad imtcp
      $InputTCPServerRun 514
      
    • Restart the service:

      sudo service rsyslog restart
      
  • Restart HAProxy service:

    sudo service haproxy restart
    
  • Monitor statistics by browsing to http://**<HAProxy server IP>**:1936/ Credentials: zadara/zadara