[pingu]

Sorry to hijack a little…

Can anybody give any hints on how the switch end should be set up for HP Procurve switches.

We have choices of ‘Trunk’ or ‘LACP’ in ‘HP speak’. As far as I can tell from the HP manual, LACP is proper 802.3ad and can be configured as either static or dynamic. Trunk is used for non-standard protocol trunks. HP recommend using Trunk when the host is using an unknown trunking protocol.

So it seems like LACP is best option, but Static or Dynamic???

Cheers
Dan

[pingu]

We finally managed to track down the problem. It looks like it was a mis-configured trunk between one of the perimeter switches and the core switch. We ‘are not sure exactly what was wrong with the config, but it looks like it may be that the network company only configured one end of it.

Since we disconnected three out of the four cables that made up the trunk all the errors seem to have gone away, our server seems to have stabilised and our network is a whole load faster and more reliable.

Dan

[pingu]

I see, I think I misread what you were saying. It would of course make sense that a clustered service would require a clustered filesystem.

I am still hoping that we see either a clustered filesharing technology and/or a better failover messaging system or hopefully both!

[pingu]

I think you’re right Joel, I had a quick look over the docs again this morning (I think it’s stil up on the developer site, but if not anybody wanting the PDF can mail me), it does seem to be more based around active/passive and really is just a messaging system. There’s nothing in the docs about true clustering (actv/actv).
That said he seemed quite confident about his answer (I don’t want to give his name or position here for fear of getting him in hot water).
When you say the storage is XSan based in Leopard, are you referring to there being a change in filesystem architecture?
I also happened to notice one of the slides making reference to AutoFS, but I know this is more about automounting

[pingu]

Just a quick update…
I was at an Apple presentation the other day introducing the new features of Leopard etc. I noticed on a couple of the slides that they were talking about Active/Active clustering for mail and iCal servers. WHen I asked the Apple bod, he confirmed that this would be based on clusterd. When I pressed him about clustering other services (namely AFP), he said he wasn’t allowed to say much, but that it was being worked on and provided everything went well would be announced later…
I wait…

[pingu]

hmmm…
It seems that fstab, under Tiger, may differentiate between hfs and hfsplus contrary to what appeared to be the case previously.
Still in search of a solution, I just tried:
[code]UUID=A507B9D0-5A73-34EF-9569-DB93E7EBF7D6 /Volumes/storarray01 hfsplus ro[/code]
which seems perhaps to be working. The volume mounts on the desktop as expected, but does not seem to be read-only. diskutil info reports that it is not read only, and the finder and shell both seem to think they can write to it; but when it is unmounted and remounted, any files or folders written have disappeared.
This worries me… alot.
Can anyone suggest a surefire way of confirming the volumes read-only status?

[pingu]

Just as a quick update BTW…

Even if a volume is formatted as HFS+, you must still list it in fstab as ‘hfs’. Listing it as ‘hfs+’ seems to cause diskarbitrationd and possibly others (mount etc) to ignore the line (I suspect they see it as an syntax error).

D

[pingu]

Thanks Joel.

Once we had configured zones on the switches and ensured that I/O Streamguard was disabled on the relevant ports everything seemed to work nicely. We did though have to do a full shutdown off all equipment connected to the switches before the hosts would play nice with the zones, so I’d advise anyone doing this to configure the zones with all other equipment shutdown.

D

[pingu]

I answer to my own question, it does seem that multipathing was included at some point.
We hooked up our two Qlogic 5200s and connected them to each other to create a cascade. The ports connecting the switches were configured as G_PORTs and the switches then auto-conf’d them as E_PORTs.
We then connected all the hosts Port 0 to one switch and Port 1 to the other, along with all out storage and libraries etc. When we boot the hosts they seem happy, and only mount one instance of each of the storage devices.
Now to get some big transfers going and start pulling out cables!
More to follow…
Dan

[pingu]

Thanks guys. I seemed to remember that it had a method of determining ‘closeness’.

Any ideas on the automounts?

[pingu]

I’ve done a bit of digging into this and it looks like using ‘xx’ actually means that diskarbitrationd and others should completely ignore the record in fstab rather than not mount it.

man fstab states: –
[quote]
If fs_type is specified as “xx” the entry is ignored. This is useful to show disk partitions which are currently unused.
[/quote]

Have you tried
[code]LABEL=XserveRAID none hfs rw,noauto[/code]

This worked for us.

Dan

[pingu]

Yeah, I think this might be a good place to start cutting our teeth. Alsothought we might give Nitrobit’s Group Policy Editor a whirl; has anybody used/tried this? How does it shape up as a solution for a smattering of Windows machines rather than trying to implement a whole AD/OD malarkey?

Otherwise, I think we may just have to ride out the wait for Samba 4…

[pingu]

I think you’re right Josh, way to delicate for production use anyways. So we’re more than likely going to go with Kerio rather than Exchange (thank god). Pricing seems more sensible too for a smallish shop.

We’d still like to provide Group Policy to the 10 percent or so (and rising slowly) of our users who we have to have on Windows (all 2k or XP). We already have these PCs bound to OD, which seems to work okay (bar a few glitches) but having to set Policy on each machine is clumsy and a lot of work.

Any ideas?

[pingu]

Hi,

Has anyone had any luck figuring out a way to do this? Is it in fact possible?

We are like carneym looking to put an Exchange server in, but we don’t want to go with a full AD implementation. We have an existing OD implementation, which works just fine. Most of our clients are Macs, with a few (horses for courses) 2K and XP machines in the mix. We just really want to be able to provide the Exchange servers with whatever it is they need from AD and the PC clients perhaps with some Group Policy, but still have all usernames, passwords etc provided by OD.

Any ideas?

Dan

[Author]

Posts