Basic hast setup questions

[daffy]

1. Does it matter if the nas4free servers run differnt versions, for example 9.2.0.1 and 10.2.0.2?
2. The "Path" in Services->HAST->Add->Path : This could be a zvol? If so, does it matter if the zvol is "upgraded" on one server and not yet upgraded on the other (got some message about zvol upgrading on a server I'm testing 10.2.0.2 on)? (I would assume this doesn't matter to HAST).
3. Does adding a device in Services->HAST->Add->Path alter the current content of that device?
4. "Node A settings" and "Node B settings". Is that different hast devices or something you do for every device? (I simply don't understand the overall basic idea)

(As you see I have some pretty basic questions.. The problem when reading links / guides on the net is that it's hard to know if information you find is outdated or current when you have very little knowledge in advance.) If someone experienced had some spare time to start a thread with basic overview information about setting up hast with nas4free I think that would be appreciated by more people than me coming completely fresh into the hast world.

[daffy]

1. Does it matter if the nas4free servers run differnt versions, for example 9.2.0.1 and 10.2.0.2?

It seems it works with different versions.

2. The "Path" in Services->HAST->Add->Path : This could be a zvol? If so, does it matter if the zvol is "upgraded" on one server and not yet upgraded
on the other (got some message about zvol upgrading on a server I'm testing 10.2.0.2 on)? (I would assume this doesn't matter to HAST).

If you want to add a zvol, you seem to need to add /dev/zvol/ to the path. (in my case I use /dev/zvol/Pool1/hast_test )

3. Does adding a device in Services->HAST->Add->Path alter the current content of that device?

No, I don't think it does.

4. "Node A settings" and "Node B settings". Is that different hast devices or something you do for every device? (I simply don't understand the overall basic idea)

It's for letting HAST know the actual location of the mirrored nodes. I got fooled by misreading "A" and "B" in the IP Address field. Make sure you enter node B's IP address in Node A's settings.

[daffy]

So, now it looks like I got HAST running.

I got 2 quite different servers with completely different hardware setups, so I believe creating a hast device out of zvols on the different servers would suit my need best.

Server 1 is running Nas4free 9.2.0.1 (full)
Server 2 is running Nas4free 10.2.0.2 (embedded)

I've created a 1Gb zvol on each server, called hast_test.
I've configured HAST in the nas4free GUI.
Had to manually run 'hastctl role [primary/secondary] ha0' from shell on server 1/2 to get out of "init".

On my primary node (server 1) I get:
hastctl status
Name Status Role Components
ha0 complete primary /dev/zvol/Pool1/hast_test 192.168.0.193

On my secondary node (server 2) I get:
hastctl status
Name Status Role Components
ha0 complete secondary /dev/zvol/pool1/hast_test 192.168.0.192

So far so good, right?

Now I want to create an iscsi target using the hast device, but haven't figured out how yet. Feel free to show me the right direction to go, but I'll get back with questions if I get into troubles googling it.

[mdsystems]

Sounds like you have worked most of it out. Nice

I guess the only comment I have about creating the HAST devices on ZFS devices is that there is more "middleware" sitting between HAST and the underlying disks. I don't imagine there is anything wrong with it, but just food for thought. I went down the route of creating a HAST device, then the ZFS vdev used the HAST devices.

As I see it, I have iSCSI -> ZFS -> HAST. You potentially have iSCSI -> ZFS -> HAST -> ZFS. I don't know how this translates performance-wise.

On the primary, if you use:

Code: Select all

hastctl list ha0

It will also give you an idea about what blocks are dirty (ie needing to be replicated) and other useful stats.

Now that you have the HAST devices, you can create your ZFS vdevs and work from there. My set-up was four HAST devices (hast0 - hast3) in a double-parity raidz configuration.

When the vdevs have been created, you create zfs volumes and these can then be assigned to iSCSI targets (first you have to create an extent that uses the zfs volume(s) that you created and these are then assigned to a target (each target supporting a number of LUNs.)

I would also check the versions HAST used by both versions of nas4Free, just to make sure that there are no known problems with the older version.

If you need any specific detail, let me know.

[daffy]

Sounds like you have worked most of it out. Nice

I guess the only comment I have about creating the HAST devices on ZFS devices is that there is more "middleware" sitting between HAST and the underlying disks. I don't imagine there is anything wrong with it, but just food for thought. I went down the route of creating a HAST device, then the ZFS vdev used the HAST devices.

As I see it, I have iSCSI -> ZFS -> HAST. You potentially have iSCSI -> ZFS -> HAST -> ZFS. I don't know how this translates performance-wise.

Hmmm.. I though I would have:
Server 1->Physical drives->zfs volume->HAST device->
------------------------------------------------(HAST "mirror")->iSCSI->"User system file system"
Server 2->Physical drives->zfs volume->HAST device->

(But as you'll see below, I might have misunderstood how to set up the whole thing)

On the primary, if you use:

Code: Select all

hastctl list ha0

It will also give you an idea about what blocks are dirty (ie needing to be replicated) and otehr useful stats.

Now that you have the HAST devices, you can create your ZFS vdevs and work from there. My set-up was four HAST devices (hast0 - hast3) in a double-parity raidz configuration.

When the vdevs have been created, you create zfs volumes and these can then be assigned to iSCSI targets (first you have to create an extent that uses the zfs volume(s) that you created and these are then assigned to a target (each target supporting a number of LUNs.)

Hmmm.. I only set up 1 hast device, which now is synchronized between the servers (as far as I understand).
I thought that I should/could set up iSCSI to use the "bare" HAST device (so it would be available even if one of the HAST servers went down). But maybe that's not the way "it's done"?

I would also check the versions HAST used by both versions of nas4Free, just to make sure that there are no known problems with the older version.

Ok!

If you need any specific detail, let me know.

Great! (For the moment I'm busy with other work, but will hopefully get back to the hast setup testing soon.)
One thing that confuse me is that if I enable Services->HAST in nas4free (9.2.0.1), the Services->iSCSI Target automatically gets disabled. Is that to be expected?

[mdsystems]

Hmmm.. I though I would have:
Server 1->Physical drives->zfs volume->HAST device->
------------------------------------------------(HAST "mirror")->iSCSI->"User system file system"
Server 2->Physical drives->zfs volume->HAST device->

That should work if you intend presenting the entire HAST device as a single iSCSI device. I was assuming that you may want to create multiple ISCSI devices, which is not necessary by any means.

Hmmm.. I only set up 1 hast device, which now is synchronized between the servers (as far as I understand).
I thought that I should/could set up iSCSI to use the "bare" HAST device (so it would be available even if one of the HAST servers went down). But maybe that's not the way "it's done"?

Looking at the UI, it is possible. I went down the route of using zfs volumes for my own reasons.

One thing that confuse me is that if I enable Services->HAST in nas4free (9.2.0.1), the Services->iSCSI Target automatically gets disabled. Is that to be expected?

I am using the latest version, so not sure, but I cannot see why enabling HAST would disable iSCSI on the same server. Is there a limitation in the old version for some reason? Clearly, the server operating as the slave will not have access to the HAST device while it is in secondary mode, and the HA configuration will ensure that iSCSI on the slave is only activated if it becomes master.