Apache and SMB high availability configuration

SUSE Linux Enterprise Extensions SLES High Availability Extension
1

HI … I need configure HA with SLES11 SP2 for apache and samba services … I have read book_ha pdf but I don’t find information for apache … I’d like step-by-step for apache and samba HA… If you have this information or link, please provide it …

thanks

vanessa
vanessav@it-synergy.net

2

Hi Vanessa,

while simply providing static content via Apache / HA is a rather easy task and well covered across the net, the devil is into the details:

  • will you need to have persistent sessions across all Apache nodes?
  • is there more to it than Apache, i.e. some back-end application, probably even a Tomcat or alike?
  • do you want load balancing or simply HA?
  • how many service-providing nodes are there?
  • where does the (static) content (that your Apache is to serve) come from? Is that HA as well?
  • what other components do you need to be HA, in terms of a “complete HA” solution?
  • how do you define HA?

Regards,
Jens

3

Hi Jens … thanks for your answer … Into My enviroment I have 2 servers with apache for publish my web site. I need implement HA because this site always must be up, not load balancing.

the content came from a NAS storage (both servers see the same content) …

I need configure High Availability with all options for 2 servers, that garantize my service always available.

Greetings,

Vanessa

4

Hi Vanessa,

that garantize my service always available.

nope, impossible. :wink: “always” is a very strict term.

Just to give you a starting point, you might go the following route:

  • set up a two-node HAE cluster using SLES11 + HAE
  • define an IPAddr2 resource for the IP address of the clustered
  • define an Apache2 resource
  • add an order constraint (Apache requires IP)
  • add a location constraint (Apache requires IP on the same node)
  • configure your Apaches on the nodes
  • start heartbeat

There can (and most probably will) be tons of other configuration adaptions necessary:

  • fencing on a two-node cluster needs special care (what if both nodes start the resources, esp. the IP address?)
  • you might get away with switching only the IP address, depending on your IP stack setup on the nodes (by running Apache constantly you’ll have quicker fail-overs, but Apache might need the desired IP address visible on the node to start up properly and bind to that IP)
  • you still have a possible SPOF (NAS) or two… (network between NAS and servers, between servers and clients, power, …)

Seems I misplaced my “building a complete cluster in 5 steps” book :wink: but in the combination of manuals and all those implementation notes out there, you ought to find your way. And we’ll be here for specific question.

Just keep in mind: HA is a design principle, not something you set up by installing two servers. That’d be like trying to win a Formula 1 race with one of those British kit cars… but those Brits will take you from A to B rather stylishly, so if that’s what you need, why not go for it?

Regards,
Jens