One of my little projects that I am at present trying to complete is a deployment of a netboot image of Tiger (10.4.11) with Office and Photoshop (CS) included to a series of Powerbook and Powermac systems.
I was thinking it would be neat to deploy them all as diskless systems, and while that works flawlessly, there is one small problem: Scratch disk space.
I was planning on using a DAS XRAID with two, 3.5TB (RAID-0) arrays as a network swap area, for scratch files from multiple systems. The problem is that in their infinite wisdom, Adobe did not encode Photoshop CS to see AFP devices as valid targets.
For whatever reason, Photoshop CS does not see the network drives at all, when I mount them in the Finder, and in fact, it doesn't see the RAM Disk either, that is generated by the Diskless boot in place of a shadow disk area.
I am wondering if anyone can offer a valid suggestion on how to trick the OS into providing these as valid targets, whether by editing mountlists to automount them and include them as targets.
I was thinking that if I can't use the network shared volume directly, that I can create R/W disk images on it that I can use per machine. The problem with that is that scratch cache clearing would become troublesome (duplicating, and then deleting old virtual volumes) and that it wastes resources, unless I use sparse image format disk images, which may work as a viable solution.
I am also concerned that even if I do use one network volume as a whole, that multiple Photoshop connections to it would conflict with each-other. I'm not sure what conventions Adobe uses for creating swap files, and if they are always unique per session, or if they always use the same identities.
Does anyone have suggestions on this kind of topic?
If you are wondering, this is for a not-for-profit group, benefiting writers and artists who are down on their luck, just starting out, or otherwise in need of assistance, that provides a cooperative work environment at no cost to them.
I have a great many spare Powerbook, iBook, Imac and G4/G5 systems that have faulty hard drives, so this is a nice way to deploy them for use without needing to put a drive and an OS on each. I'm talking about a dozen or so design stations up at any tone time at maximum, not hundreds, so I'm not terribly concerned about bandwidth, especially as some of the systems don't even have 1000bT Ethernet, although most do.
I'm not even sure why the RAM Disk and the AFP volume aren't valid scratch targets. I have not yet tried the disk image method to see if that would become a valid target, but it is next on my list.
Update: Problem semi-solved. I created a series of sparse images on the main file node, and when mounted, PS CS can use them as swap. I still don't have a clear reason for what it can't do this with an AFP mounted volume (on 10.4), and I am still curious on how it names, allocates and protects swap files on the swap (scratch) volumes, if at all.
If anyone else ever has a similar issue, sparse disk images are your solution. If you only run OSX 10.5 and later, then you can alternatively use sparse bundles; however in my situation with 10.4, 10.5 and possibly 10.3 systems, the standard sparse image is vital.
I should also note that when I created a sparse image on 10.5, 10.4 couldn't read the filesystem, however when I used 10.4 to create it on the network volume, 10.4 could read it, and 10.5 could not, so your sparse images (not bundles) are OSX specific before and after 10.5 for whatever reason. I still need to check if a 10.4 sparse image will mount in 10.3, but I have a feeling that it will, and that 10.5 is the breakpoint because of Time Machine support.
I was thinking it would be neat to deploy them all as diskless systems, and while that works flawlessly, there is one small problem: Scratch disk space.
I was planning on using a DAS XRAID with two, 3.5TB (RAID-0) arrays as a network swap area, for scratch files from multiple systems. The problem is that in their infinite wisdom, Adobe did not encode Photoshop CS to see AFP devices as valid targets.
For whatever reason, Photoshop CS does not see the network drives at all, when I mount them in the Finder, and in fact, it doesn't see the RAM Disk either, that is generated by the Diskless boot in place of a shadow disk area.
I am wondering if anyone can offer a valid suggestion on how to trick the OS into providing these as valid targets, whether by editing mountlists to automount them and include them as targets.
I was thinking that if I can't use the network shared volume directly, that I can create R/W disk images on it that I can use per machine. The problem with that is that scratch cache clearing would become troublesome (duplicating, and then deleting old virtual volumes) and that it wastes resources, unless I use sparse image format disk images, which may work as a viable solution.
I am also concerned that even if I do use one network volume as a whole, that multiple Photoshop connections to it would conflict with each-other. I'm not sure what conventions Adobe uses for creating swap files, and if they are always unique per session, or if they always use the same identities.
Does anyone have suggestions on this kind of topic?
If you are wondering, this is for a not-for-profit group, benefiting writers and artists who are down on their luck, just starting out, or otherwise in need of assistance, that provides a cooperative work environment at no cost to them.
I have a great many spare Powerbook, iBook, Imac and G4/G5 systems that have faulty hard drives, so this is a nice way to deploy them for use without needing to put a drive and an OS on each. I'm talking about a dozen or so design stations up at any tone time at maximum, not hundreds, so I'm not terribly concerned about bandwidth, especially as some of the systems don't even have 1000bT Ethernet, although most do.
I'm not even sure why the RAM Disk and the AFP volume aren't valid scratch targets. I have not yet tried the disk image method to see if that would become a valid target, but it is next on my list.
Update: Problem semi-solved. I created a series of sparse images on the main file node, and when mounted, PS CS can use them as swap. I still don't have a clear reason for what it can't do this with an AFP mounted volume (on 10.4), and I am still curious on how it names, allocates and protects swap files on the swap (scratch) volumes, if at all.
If anyone else ever has a similar issue, sparse disk images are your solution. If you only run OSX 10.5 and later, then you can alternatively use sparse bundles; however in my situation with 10.4, 10.5 and possibly 10.3 systems, the standard sparse image is vital.
I should also note that when I created a sparse image on 10.5, 10.4 couldn't read the filesystem, however when I used 10.4 to create it on the network volume, 10.4 could read it, and 10.5 could not, so your sparse images (not bundles) are OSX specific before and after 10.5 for whatever reason. I still need to check if a 10.4 sparse image will mount in 10.3, but I have a feeling that it will, and that 10.5 is the breakpoint because of Time Machine support.
Last edited: