Loadballancing plesk

[Hidden]

Hello,

i'm using plesk for some time now and i'm wondering of it is posseble to loadballance the plesk server.

i have a hardware loadbalancer, 3 plesk servers and a shared storrage.

so if one of the servers dies the websites are still online.

 

[tiramisu]

Hi,

Plesk has no such utility or option to share its' data, to create a load balancing.
what features Plesk does have are Plesk backup that allows to backup all Plesk data. In case of failure, you can restore data from it.
also you can backup all server with such application as Acronis Image.
search at acronis.com

also read docs of Plesk:
http://download1.swsoft.com/Plesk/Plesk8.2/Doc/plesk-8.2-unix-administrators-guide/index.htm

 

[Hidden]

the problem is you have to do it manual so there is always downtime if you don't use loadballancing or some thing simmilar.

it would be a great option if swsoft create this in plesk

 

[tiramisu]

Hi,

yes, that might be useful option,
but i'm afraid it will be implement not soon;(

 

[sbillis]

Just a thought as I have yet to test this - but it might be possible using DRBD to replicate the storage and configurations at a block level. I was thinking of doing the following:

Creating a pair of servers running DRBD and linux heartbeat. Have the data to be replicated stored on the DRBD device and then have the ip configurations handled by linux heartbeat.

I've yet to determine the exact method of doing this but this may provide you with some help.

Rgds

Simon.

 

[tiramisu]

another way:

1) creating mirror RAID that duplicates all the data stored; this will allow to get Plesk up and running from the reserver partition in case of hardware failure;
2) schedule 'rsync' task in crontab; this task will copy all the data to remote host and update them permanentnly; note, that remote host should be offline (but powered on though) for not to checking license via the Internet.
3) any other;

 

[joda]

I have to agree, with the other posts here...

There is NO way to make a PSA cluster, due to multiple reasons.
The best way to secure your setup, is using high-quality hardware like, IBM x346 or HP DL380G4, using harddrives setup in RAID.

If you have one client, where you are hosing one website - there is a few things you can to create a cluster like environment. This is a hole new topic, and is you need info on this let me/us know.

 

[stupidnic]

What you are looking for is not a truly load balanced application, but rather a hot stand-by configuration.

You can accomplish this using the servers, the shared storage you currently have and a virtual server technology (Xen is my recommendation) on each front end server.

You can setup a virtual server that stores the entire plesk virtual server on the shared storage (disk images for the root, swap, tmp, etc). In the event of a failure in the hardware of the primary server, you can boot the virtual server on the hot stand-by server and be back up in running in the time it takes to complete a fsck of your file system images.

Take a look at Xen and its Live Migration functionality.

I will put out a caveat though, unless your servers are using a shared back end with a high throughput (4Gb iSCSI SAN, Inifi-band, or some sort of shared direct attached storage) the performance of the virtualized server to the back end storage system can be slow, and will cause some customers to complain.

We are waiting for the price of 10GigE to come down to a reasonable level before we start deploying something like this in a production environment.

Virtualization is the next step in servers, as it completely decouples the OS from the underlying hardware. There are just a couple of things that need to occur before we are completely there, a high throughput, relatively inexpensive, back end storage system is probably the biggest hurdle.

 

[cttt]

We currently use Xen and DRBD for failover with Plesk.

The guests (virtual servers) are not load balanced, but it takes a matter of seconds for the spare to take over when there is a failure on the primary.

We have one machine acting as primary for one guest, and another machine acting as primary for another. DRBD keeps the data synchronized on both machines. If one machine fails, the backup guest starts on the other machine. Performance suffers a small amount, but other than that, the outside world has no idea what happened.

 

[naseer6679]

Documentation

Hi,
Please can any one of you provide me with a documentation of implementating failover using xen and drbd.
Thanking you.
Regards,
Naseer.