Skip to content

Latest commit

 

History

History
114 lines (85 loc) · 4.62 KB

7-general_design_issues.md

File metadata and controls

114 lines (85 loc) · 4.62 KB
Raw

General Design Issues

Networking

Multihomed servers

  • a server having more than one NIC
  • multilple prod + backup/admin network
  • bonding= two NIc sharing the same IP (careful with routing loops)
  • By default, an application that listens on a socket will listen for connection attempts on any interface
  • configurable properties to define to which interfaces the server should bind

Routing

  • with multiple interface it can get tricky
  • keep track of destination name, address, desired route for each integration point

Virtual IPs

  • two meanings:
  • Load balancers use virtual IPs to multiplex many services (each with its own IP address) onto a smaller number of physical interfaces
  • Generally speaking, it means an IP address that is not strictly tied to an Ethernet MAC address

Security

The Principle of Least Privilege

  • any app -> lowest level of privilege needed to accomplish its task
  • webserver behind loadbalancer listening on port greater than 1024
  • <1024 apache uses privilige separaten (gives up root once socket is open)

Configured Passwords

  • prod passwords separately.

Availability

Gathering Availability Requirements

  • actual cost vs. avoided losses

Documenting Availability Requirements

  • SLAs in terms of specific features or functions of the system
  • SLA can't be greater than the worst of the external dependencies involved in a feature
  • SLA should define what device or devices will be monitoring the availability of the feature (e.g by executing synthetic transactions)
  • checked how often?
  • acceptable response time?
  • from how many locations?
  • how will it be recorded?

Load balancing

  • Load balancing is all about distributing requests across a pool or farm of servers to serve all requests correctly in the shortest feasible time

DNS Round-Robin

  • simply associates several IP addresses with the service name
  • drawback: all the servers in the pool must be “routable". (their front-end IP addresses are visible and reachable from clients) -> security risks
  • drawback2: putting too much control in the client’s hands -> no way to remove host from the cluster, can't guarantee even distribution
  • has to bypass Java DNS caching if used for backend services

Reverse proxy

  • A normal proxy multiplexes many outgoing calls into a single source IP address.
  • A reverse proxy server does the opposite: it demultiplexes calls coming in to a single IP address and fans them out to multiple addresses.
  • does its magic at the application layer. As such, they are not transparent
  • can be configured to cache static content.
  • the biggest reverse proxy server “cluster” in the world is Akamai. (a large number of servers located near the end users,)
  • Because the reverse proxy server is involved in every request, it can get burdened very quickly
  • keep track of origin health
  • X-Forwarded For header

Hardware Load Balancer

  • expensive

Clustering

Load balancing does not require collaboration between the separate servers. When the servers are aware of each other and actively participate in distributing load, then they form a cluster. some communication overhead in heartbeats and state synchronization

  • active/active - loadbalanced, all serving traffic
  • active/passive - redundancy, in case of failure
  • WebLogic, MSSQL has it built-in

Administartion

  • easy to administer, it will have good uptime
  • users and devs - releases add value, ask admins/ops
  • new release, new failure modes, old logs might disappear
  • make your ops guys happy, DEVOPS

"Does QA Match Production?"

  • why the issue wasn’t discovered during testing?
  • guaranteed to find something different (hostnames and IPs at the very least)
  • the expensive part is figuring out which discrepancies are known, expected, and harmless and which are unexpected and risky
  • tipcally confgiuration and topology differences
  • Topology is the number and connectivity of the servers and applications
  • sharing hosts that'll be different in prod. (solution: virtualize in QA)
  • Zero, One, Many - if prod has twelve run at least two in QA

Configuration Files

  • enterprise system: loads of configuration files
  • config paramater names has to be unambigous
  • internal config (bean wiring) vs application config (hostnames)
  • put your shit in version control
  • configuration properties are part of the system’s user interface

Start-up and Shutdown

  • logged startup with logging errors
  • start-up sequence must complete successfully before the application starts accepting work
  • fail fast
  • clean shutdown, waiting for in-flight transactions (with timeout) but not accepting new work

Administrative Interfaces

  • if GUI provide scriptable alternative