Thursday, March 4, 2010

Complexity of LH

This post (although take with a grain of salt) points to possible more complexity and lock ups for LH.

Sorry, you will have to in order to view it

A couple of out takes

Theire units work well for the most part, but we've had far more issues over the past years with the units locking up than you'd expect, almost always due to an incompatabilities to their OS and the underlyring RAID controller - I belive at this point we've had to upgrade the controller on 3 different occasions in 2 years. The feature set from Lefthand is quite nice for the price, but its missing a couple things, like being able to manually move the cluster VIP between boxes (its an automatic thing, so iy only does it when you down a box).

Performance on the units is fairly good, but make sure you calculate what IOPS you'll get from the units versus what you'll need to run your servers and VMs off of it. Both vendros can give estimates on what the units will give you, and what sort of number they'd expect in your environment.


Not a ton (we're still using them), but more than I feel we should have. The units don't failover as gracefully as they should when one has a problem, and so we've had to do do a fair number of manual work when there's an issue to get things back running.

Example: at our HQ, we're using 2 mirrored units to host a number of mid to high importance servers - a couple VMs, data store to a couple important but low use linux servers, and a fileserver that hosts the folder redirection to laptops. When one unit hands/locks up, its supposed to gracefully fail over to the other unit, with no loss or connection drops. What actually happens is the fileserver dismounts the drive, and both the VMs and the data store file systems go to a read only state. It's all fixable, but causes significant scrambling to fix the problem.

squigit, there are two ways to set up network RAID on LeftHand., in one mode it stops if a node goes down for a high level of data protection, in the other mode it keeps going. You can also migrate a LUN to another set of nodes by jiggling about with cluster membership. Maybe you just to install the failover manager on a 3rd box to maintain quorum. Sounds like you need to go on the install/config course.

No comments:

Post a Comment