lspeed.org

Apologies for the lack of updates on here. Many changes are in progress and I suspect this site will remain quiet for the time being. But I'll take this opportunity to provide some updates...

1) I bought a 36U rack on Ebay! It is totally awesome and I can't wait to start using it properly. However that will have to wait for number (2). It is an HP 10636 G2 rack, and virtually as-new. Currently being stored in the kitchen.

2) I hope to be moving to a new flat within the next 2 months. Only 200m away from the current place, but this time I'll be making mortgage rather than rent payments! Anyway, I am not yet entirely certain that the new rack will fit up the stairs to the new place, but fingers crossed that I find a way to get it in!

3) Windows is getting on my nerves and I have serious plans in motion to replace it with linux (Gentoo) almost throughout the network. Central to getting this working is migrating server functionality between the platforms. Currently I have 4 virtual machines running inside the Opteron box, on which I am setting up the core functionality on what will later become the 4 real phyical linux servers. The basic idea is that there will be two front end x86 SMP boxes running services that are accessible from the outside world, and two backend x86_64 SMP boxes running everything else, including a SAN. Having two of everything means I can play with clustering and I intend to do both loadbalancing and high availability for as many of the services as possible. This is obviously a major project and I suspect it will take most of the year to complete :)

So far I have been concentrating on the storage side of things, since most other stuff will depend on that being in place already. The overall design for the SAN is as follows: Use EVMS2 to allow flexible allocation of the underlying disk space, on top of which there will be an active/active DRBD v8 setup on the SAN boxes, with both DRBD nodes exporting the devices as iSCSI targets. All the servers (including the SAN boxes themselves) will then connect to the iSCSI targets on both SAN boxes, and dm-multipath will take care of the failover stuff. CLVM and GFS2 will be used to provide a cluster-aware shared storage filesystem that can then finally be mounted and accessed by all nodes simultaneously. This will all run across multiple bonded gigabit network links to try to minimise the performance impact as much as possible.

4) With all of this to keep me occupied I have neglected gaming somewhat so far this year :( but not completely! STALKER was great fun, and now completed. Hurry up Crytek and release Crysis! 

Tags: