Surprised to find this blog still here. It’s been…oh…a long time since I’ve ventured into the blogging world. Work has kept me busy…going into year 4 of a six month contract and making all sorts of discoveries of late.
Discovery #1 – Brocade is still a third-rate switch company. The hardware is fairly bulletproof, when it comes to reliability… But they’re still married to the idea of “local-switching” as an alternative to building a backplane that’s worth a damn. Sorry, I’ll take the Cisco MDS 9700 series any day of the week and twice on Sunday.
Discovery #2 – Well not a discovery really. EMC Symmetrix (Symmetrix/VMAX) is still the flagship storage array. If you put anything else in, you’re going cheap. Not that there’s anything wrong with that, but it’s time to admit that that’s what you’re doing.
I say that having worked now with HP 3par – which I put as equivalent to the Clariion/VNX line in stature and performance, and HP XP7 (Hitachi G1000) which is higher end, but a bloody nightmare to manage. (I don’t know if that’s Hitachi or just HP’s version of Hitachi that makes it a nightmare, I’ll have to wait until I get hands on an actual Hitachi to see)
Let me be clear, this is a personal preference. Both arrays, 3par to some extent, and XP7 to greater extent, seem to be trying to steer people away from using the CLI to manage their arrays. GUI’s are fine, but they don’t offer the level of control that you need to micromanage the hell out of your storage (as I like to) And GUI’s also make scripting changes more difficult, and more prone to error.
I haven’t had a chance to really beat the daylights out of the XP7 yet but will in the months to come, I’ll report further as I discover.
The HP XP7 IS the HDS VSP G1000. They are the same.
HDS always has “interesting” management interfaces. Other than that the storage itself is rock solid and we never had any issue with them.
We don’t have any G1000 here but our VSP can be managed by CLI through HORCM. It is supposed to be easier and more comprehensive now with the new G series.
When I took the class, they told me specifically that there are a number of functions that can’t be done via HORCM….has to be done via the GUI.
I’ve found that virtualization of new devices is one of those, or at least I haven’t found a way to do it via the command line.. Point-and-click-only so far…which is tedious as hell.
Sorry it took so long to reply… I keep forgetting about this. 🙂
The CCI manual showed commands for external volume. I suppose those are for virtualization?
We have never tried those as they are G1000 specific. The command set usable on our VSP is much smaller.
LOL god no.
External volumes are formatted volumes to be used as XP7 Back-End devices. Data is not preserved when converting a volume to external (found this out the hard-way on our test array thankfully)
Virtualized volumes are pass-through pseudodevices. (data kept intact) It’s a whole different process to virtualize volumes.
HORCM: Replication (Internal, and external), and SnapShots
RAIDCOM: Provisioning. Anything you can do via the GUI, you can do using raidcom.
CVAE: hdvmcli.sh, data collection.
I use all three, every day. I cannot find anything I cannot do via the command line.
Maybe the problem I’m having is I haven’t gone into depth on the CVAE command line. Right now the process is, provision via command-line, and then virtualize and migrate using the GUI. Its kludgy at best and anything point-and-click is subject to errors so you have to be extra paranoid. (Anything CLI is subject to errors, but I’m a better typist than I am a mouser)
It’s the usual learning-curve on a new technology. So far I’ve found the hardware to be pretty solid, but I think the GUI was built by monkeys on acid… It’s clumsy and not very intuitive.
The problem with doing data-collection via CVAE is that the database has a very real chance of not being up-to-date. So if I build a provisioning script based on data I get from CVAE, I could easily step on LDEV’s that have already been attached, etc.
I long for the instant response of SYMAPI… (I also long for the one-command-set-fits-all nature of it, oh and not having to have a separate server, and a myriad of other things.)