High Availability

[dgitman]

I was wondering if anyone has come up with a good high availability (SLA) solution. Is anyone running a active/active or active/passive cluster? I'm shooting for an active/passive, since they're generally less complex. The only scenario i could come up with would be to rsyc (hourly) the home directories and use the replication built into MySQL.

cPanel.net Support Ticket Number:

 

[ssilvius]

This is somthing we are working really hard to achive. We are writing some code that will mirror a single install over a farm and using BIGIP to manage the farm connections. Once you start thinking about HA, you have to spend money, it's a simple as that.

We're going to be moving from the pile of 1u boxes to dell blade servers using BIGIP for blades in the "master" enclosure.

We are testing two things right now. The code mentioned above that acts as a replication mechinsim. The second is using a shared storage array between all the blades. The second is a SAN on giganet copper so it's very fast. The blades all have two network connections and we are looking at using one of them for storage only and the other as the blades connection to the outside world.

Personally the first one seems counter productive to me because we are still limited to a single "master" . Hardware seems to be the best choice for us, but I trust my crew and if they say that they can make their replication system safe and fullproof, well let's see.

cPanel.net Support Ticket Number:

 

[dgitman]

What kind of storage can you use with the blade servers? Only SAN? What about dealing with SQL quires and fail over. I don't think that would work well.

cPanel.net Support Ticket Number:

 

[ssilvius]

The SAN is raid 50, overly redundant. only the /home directories are hosted by the SAN.

The blades will each have raid 1 within for the system software.

What failover? BIGIP manages the farm, and that is redundant also.

The "master" blade enclosure will be two dns, two firewall, two BIGIP.

The "farm" blade enclosure will be six webservers. These do apache nothing more.

if an apache blade dies, oh well it gets yanked and fixed. BIGIP sees it's dead and does not send traffic to it, the other five keep serving.

mail and sql servers for now stay where they are at, on their own machines.


edit// we may use mail on the apache blades. Still looking at doing that as it would be easier to not have to hack cPanel.
cPanel.net Support Ticket Number: