PDA

View Full Version : Booting From RAID



Louie
11-15-2000, 05:45 PM
Now that I have a RAID 0 volume and other volumes formatted with SoftRaid, I'm having the previously reported problem of having to use System Picker to change boot volumes. All the SoftRaid driven volumes appear highlighted in the Startup Disk Control Panel.

Now I discover that the RAID volume freezes on bootup as soon as the backside cache is activated. The only way that I can boot is by checking "Write Through" in the XLR8 Control Panel. Never needed that before. What's happening?

magician
11-16-2000, 12:47 AM
I haven't encountered that problem, Louie. Let's see what K says.

magician
11-16-2000, 12:49 AM
oh, yeah.

I spoke to Mark James about that bug, and he indicated that it would be fixed in MacOS 9.1, aka Fortissimo.

I believe this is now slated for release around Jan, probably MacWorld.

Ton
11-16-2000, 06:27 AM
Louie,
exactly the same symptomes I have.
Although system picker and "write through" are working perfectly for me and also I don't see any noticable speed penalty not using the write-back option.
Cheers

jorge
11-16-2000, 10:54 AM
Darin encountered this issue as well. I really haven't since I have pulled all my drives off the internal bus.

Happend on my brother's http://macgurus.com/infopop/emoticons/icon_frown.gif 9600 (boy do I miss it) when I reinstalled the 4Gb. Just pulled the system folder off the drive. The drive just authorize all the audio program since they require a drive at SCSI 0.

j.

Louie
11-16-2000, 12:32 PM
I think Kaye is off on something directed by the House Goddess. He now has to use System Picker since he switched to 9.04. He does not have to use Write-Through.

Can anyone explain "Write-Through"?

It doesn't slow the ATTO results. I'm still getting 76 SR and 70 SW.

[This message has been edited by Louie (edited 16 November 2000).]

magician
11-16-2000, 05:21 PM
there are white papers on the XLR8 site that explain cache flushing and the "write-thru" option.

kaye
11-16-2000, 06:49 PM
I'm here. House Goddess has released me for a few minutes though my email this morning had several people asking about Gurus questions.

As soon as I started using OS9.0.4 the Startup Disk CP started showing the problem you descrbe with System Picker the solution. If I boot off of 7.6.1 or 8.6, the Startup Disk CP works fine. Any OS9.0.4, whether on RAID or not, whether on Miles2 or not, even on my original 50-pin HD on the internal chain which has the latest Apple driver and not SoftRAID, has the same problem with the Startup Disk CP.

Write-Through description from Breedon's site for the xlr8 G3/500:

"The Advanced settings page will allow you to set the backside cache speed to manual (you select a speed) or automatic (the default based on the card type and bootup testing). At cache speeds above 250MHz, the new 'write-through' cache setting becomes enabled as a safety precaution, since the faster copy-back (default mode) may not be reliable. In both the Genesis and 9600/350, I was able to uncheck this and run copy-back mode at cache speeds of 262.9 MHz (when testing at CPU speeds of 525.8MHz (525.8/262/9/52.6). Results in a different Mac with another sample may not be reliable at these speeds however."

"Copy-Back (default) vs Write-Through Cache Modes: The default 'copy-back' cache mode is faster as writes are written to the backside cache (and later copied back to RAM if data has been changed). Write-through mode means that each CPU write is written directly to much slower main system RAM. Interesting is that the B&W G3s are set for write-through modes, but with their faster 100MHz memory bus speeds it's less of a performance issue than with older macs with much slower memory bus speeds. The 'Write-Through' option only becomes visible at cache speeds of over 250MHz (i.e. - when running at CPU speeds of over 500MHz)."

"Although I do not recommended it to others, I was able to uncheck Write-Through mode and still run reliably at the maximum speed this CPU card in these systems would run - 525.8/262.9/52.6. I saw no errors in extensive tests at this speed, but these results may not be repeatable and are not recommended to others."

"Speculative Processing:
There is also an option to enable or disable (default) Speculative Processing of the G3 CPU. Speculative Processing is often called 'branch prediction' - where the CPU tries to predict the execution paths of the program code. As noted previously, older Macs do not have 'G3-Clean' ROMs so it is advised to keep Speculative Processing disabled (required for software like Retrospect Backup to work 100% reliably). There is a small (4% or so) performance hit in many applications with Spec. Processing disabled, but Photoshop 5 tests actually showed higher filter performance with it disabled in my tests."

Now, what does Write-Through enabled do to my PTP with xlr8 G3/500 running 540/270 in OS9.0.4? I ran tests both ways in rebooting after each in MB5:

No/YesW-T----Proc---FP---Graphics---PubGraphics----Disk/PubDisk
-noW-T-------1785--2289----3912-------2713-----------yes or no, no change
yesW-T-------1658--2119----3677-------2645-----------yes or no, no change
yesW-Tchange-(-7%)-(-7%)---(-6%)------(-2.5%)

Disk and Pub Disk have to go thru main memory anyway, so I would not expect a change. But Processor, FPU, and graphics go thru the Processor for crunching so they are affected by the main memory of Write-Through. I also tested Memory Throughput with Gauge Pro. No change with Write-Through. It is just the use of main memory with Write-Through that causes the slowdown of certain operations.

Have I left anything out? k

magician
11-16-2000, 07:30 PM
hmmm...i just surfed thru the XLR8 site and couldn't find shineola about "write-thru" mode.

I'll check with Darin....and look around on my machine. I have a bunch of white papers they have sent me over the years....they just aren't filed as neatly as you would think.

kaye
11-16-2000, 07:36 PM
The discussion of Write-Through mode I found below was within his review of the xlr8 G3/500 on the software controls page. k

Louie
11-16-2000, 09:21 PM
OK. Thanks Kaye. Now I know what it is.

Does your machine default to Write-Through since your backside cache speed is over 250? Obviously it will run either way.

The question is; why does my machine suddenly need Write-Through to boot with backside cache enabled after switching to SoftRAID and striping two drives? (It will boot with the cache disabled.) My backside cache speed is 240, as it was before.

kaye
11-16-2000, 09:50 PM
My machine does not default to Write-Through, it defaults to showing the Write-Through check box above a backside cache of 250. It will default to Write-Through if at the beginning of bootup it does that test with the thermometer bar, or if after bootup and at the desktop it opens the control panel and asks if I want to run a slightly faster speed. Otherwise, normal boot at 540/270 Write-Through stays disabled. If I boot with the card's switches set for 500/250, the Write-Through check box is not there.

"The question is; why does my machine suddenly need Write-Through to boot with backside cache enabled after switching to SoftRAID and striping two drives? (It will boot with the cache disabled.) My backside cache speed is 240, as it was before."

The info from Breedon's site was from his test of the xlr8 G3/500 card. My guess is that your card, xlr8 G3466 card?, will cause the control panel's Write-Through check box to show anytime you overclock, or any xlr8 card is overclocked. Now to your RAID. Striped RAID really stresses the whole system and I'm guessing that the backside cache just can't handle it.

As I mentioned in my email, you might be better backing off the overclock rather than accepting the percentage hit I showed for certain MB5 operations. A 7% hit on my card if running 536/268 with Write-Through enabled is the equivalent of going back to 500/250 with Write-Through disabled. k

Louie
11-17-2000, 12:38 AM
That all makes sense. I wish Darin would jump in here and tell us the straight skinny from the Wizards in GA.

One thing Kaye; my "Write-Through" check box doesn't appear unless it's been checked earlier. I had to click on "Advanced" with the OPT key held down in order for it to show the first time.

kaye
11-17-2000, 02:13 AM
Well that blows my theory about Write-Through check box appearing anytime an xlr8 card is overclocked on backside cache. As you said in your email, it must be backside cache above 250. k

[This message has been edited by kaye (edited 17 November 2000).]

Louie
11-17-2000, 02:38 AM
Here's really weird one. I backed it off to 466/233/46.6 and it still freezes as soon as the cache is enabled unless "Write-Through" is checked. The slower speed dropped my memory performance about 5%. So, I'm back to 480/240/48 with
"Write-Through" enabled.

kaye
11-17-2000, 03:14 AM
Louie,

Looks to me like Write-Through is the only way to go then. Something about the dual-bus striped RAID the xlr8 card does not like even at stock speeds. Say, I wonder if balance is afflicting you. Pull the power plug on the 9GB 10k Cheetah which is on one of the Miles2 with the X15, whereas the other Miles2 only has an X15, right? You can leave the SCSI cable connected.

If you boot from one of your non-Miles2 drives, everything is OK at 480/240 without Write-Through needing to be enabled? k

[This message has been edited by kaye (edited 17 November 2000).]

magician
11-17-2000, 10:25 AM
it would be better to pull the SCSI cable and leave the power cable attached.

I suspect Cooksie may be the only guy who can definitively answer this, but I'm suspecting that your machine is testing cache on start-up, and deciding to enable write-thru at that time. Much of the functionality of the MAChSpeed Control control panel application is automated, and it may be detecting some condition during boot that automatically enables write-thru. This is just speculation on my part. I'll email the cowboys at XLR8 and see if they can jump in here.

Ton
11-17-2000, 10:43 AM
I have no knowledge about the differences between the XLR8 drivers and for instance the Newertech drivers. Fact is that both drivers have the "write thru" option and on my machine the problem disappears with using the XLR8 driver as well as the Newertechdriver with write thru enabled.

Louie
11-17-2000, 02:20 PM
No Magician; there's nothing "automatic" about this. From the very first time trying to boot from the RAID it would lock up as soon as XLR8 extension loaded and activated the backside cache.

Out of desperation (and reading ton's earlier posts), I forced the Write-Through option to appear by Option Clicking the Advanced tab in the XLR8 Control Panel. It never shows unless it has already activated or forced to show.

Once last night I got it to run without write-through by lowering the backside cache speed below 200. However, the CPU is rated 466/233 and it still needs write-though at that speed.

XLR8 Engineering
11-17-2000, 03:21 PM
Write-through is a slower, safer method of using the backside cache. Whenever anything is written to the cache it is automatically "written through" to main memory. This causes a slow down of about 10% in general but can also increase stability, especially with PCI cards installed.

Here are the rules for when write-through is automatically enabled:
1) When any processor is running faster than 510 MHz
2) When a G4 is installed on any machine prior to a Beige G3 (from the "clone" era)

Here are the conditions which will cause the write-through check-box to be visible:
1) Option-key transition to the Advanced Panel
2) Write-through already enabled
3) When any processor is running faster than 510 MHz
4) When a G4 is installed

Note that the write through box is always visible when a G4 is installed. The G4 being a new processor at the time of the last build, we felt it would be safer to leave that option available even on newer machines "just in case".

Hope this helps.

Chris Cooksey
Director of Engineering, XLR8, Inc.

magician
11-17-2000, 04:12 PM
damn--there's nothing like having a real authority step in and put all speculation to rest!

http://macgurus.com/infopop/emoticons/icon_smile.gif

thanks, Chris!

that helps a lot!

http://macgurus.com/infopop/emoticons/icon_biggrin.gif

Louie
11-17-2000, 05:10 PM
Thanks Chris:

Can you think of a reason why I could run the PTP at 480/240/48 for a year without even considering "Write-Through" and then be required to use it when adding two Miles 2 cards, two X15 Cheetahs, one 10k Cheetah and striping the two X15's (even when slowing to 466/233/46.6)?

Louie
11-17-2000, 11:34 PM
In addition to the RAID volume I have four other boot volumes; OS 9.04 and OS 8.6 volumes on an 18 GB UltraNarrow Barracuda hosted by a BlueNote and OS 9.04 and OS 8.6 volumes on a 9 GB 10k Cheetah that is hosted by one of the two Miles 2 cards that handle the two-bus RAID.

I am able to boot and run flawlesly from any of these four volumes at 480/240/48 without checking "Write-through". I can even run ATTO on the RAID volume fom any of these other volumes

The RAID volume refuses to boot or run unless "Write-through" is checked.

I haven't had a chance yet to deactivate the 10k Cheetah to see if that changes anything.

Darin, when you finish playing with pet mice, how about some wise comment here.

Louie
11-18-2000, 10:52 PM
I just tried removing the 10k Cheetah from the system and got exactly the same situation as described below.

kaye
11-19-2000, 11:44 AM
Sounds like we need Darin or Chris. k

magician
11-19-2000, 03:14 PM
I doubt that they will know.

This may fall into the category of things we'll never understand--fortunately things work!

Louie
11-20-2000, 12:48 PM
My $2000 problem has not gone away. Of all you RAID users, there are no other ideas?

jorge
11-20-2000, 01:59 PM
Louie,

I haven't been following your thread but have you tried pulling all the scsi devices and PCI cards except the four 15k cheetahs and booting at the default 500 Mhz? Seems like you have quite the SCSI ID collection there. Do you have a spreadsheet to keep track? (just a joke) http://macgurus.com/infopop/emoticons/icon_smile.gif

Try Darin's solution of formatting the x15s as hfs and not hfs+.

I understand what you are going through Louie. Pisses me off when B&W G3s and Sawtooths have been out for years and no one knows how to get around the raid booting in these machines. The G4 block was quickly taken care of, but the raid issue remains. http://www.macgurus.com/ubb/alert.gif

j.

[This message has been edited by jorge (edited 20 November 2000).]

magician
11-20-2000, 03:01 PM
Louie, doesn't your Mac work correctly with "write-thru" enabled?

Louie
11-20-2000, 04:33 PM
Magician:

It works "correctly" if you consider that Write-through must be activated to boot and run from the RAID volume whereas Write-through is not needed for four other boot volumes. Also, I'm not anywhere near the criteria that Chris said would necessitate Write-through.

I was never my intention to spend $2000+ and a week's work to go faster and then have to take a 7-10% performance hit with Write-through.

And Jorge; thanks for the comment. Except for Kaye's help, I've felt like a leper. This was not supposed to be a problem.

[This message has been edited by Louie (edited 20 November 2000).]

Ton
11-20-2000, 04:53 PM
Hey Louie, I understand the way you are feeling about this.
Although maybe I'm a bit less disappointed than you about having to run with write thru enabled I can't get this thing out of my head and I keep tinkering to get it working as its supposed to do.

Louie
11-20-2000, 05:08 PM
Thanks ton:

I was about to do a search and reread your earlier posts on the matter.

magician
11-20-2000, 11:33 PM
sorry you feel like a leper, Louie.

I have no clue why it would behave like that. I don't think anyone knows.

Louie
11-20-2000, 11:58 PM
I just went back and read Ton's thread from Aug-Oct. It's easy now in retrospect to see that his prolonged agony was being triggered by exactly what I'm seeing. The instant that the back-side cache kicks in at startup, everything freezes. Enabling Write-through is a workaround.

In Ton's case, the Newer card was blamed. But, I have an XLR8 card. This points me to SoftRAID 2.2.2 or the Miles 2 card/firmware.

kaye
11-21-2000, 12:28 AM
Louie,

Did I ask you in email whether your RAID only has 9.0.4 on it? My memory sucks, afflicted with speculative processing. Have you tried booting with 8.6 from the RAID? k

Louie
11-21-2000, 12:34 AM
You asked me, but I can't remember if I answered.

My RAID was My Step One; one huge volume, no partitions, OS 9.04 only. So, I haven't tried booting from 8.6 on the RAID volume. It boots and runs great with 8.6 and 9.04 on other volumes including the 10k Cheetah on one of the Miles 2 cards.

[This message has been edited by Louie (edited 20 November 2000).]

Ton
11-21-2000, 03:49 AM
My feeling, but thats just what is, is that the raid is stressing the bus to much, which could point to Softraid or The Miles firmware as Louie expressed before.

magician
11-21-2000, 06:15 PM
what happens when you run off just one Miles2, not two?

if you think about it, there could be something very, very low-level happening when you bus duplex that forces the backside cache to get flushed.

Louie
12-04-2000, 11:47 PM
In my continuing quest to solve the "Write-Through" requirement which shouldn't be a requirement, Kaye loaned me some of his toys to play with.

He sent a brand-new, matched gift set of 128 MB FPM RAM. Seriously, these were wrapped together; a brace if you will. Instead of being either 16244 or 162244, this stuff was 163244AA. It had stickers that said 60ns, but the chips were all marked -5, indicating 50ns to me.

I pulled all of my RAM and it replaced it with Kaye's. No change. Incidently, mine is all 162244, 60ns.

Kaye also sent his XLR8 Carrier with 500 MHz CPU. This is the one he has been using in his PTP tests and it was still set at 540/270/54!

I pulled my Carrier with a 466 that has been running forever at 480/240/48 and replaced it with Kaye's card. (Yes, I set it back to Default of 500/250/50; I didn't want it to melt on my watch!)

Result; no change!

Next, I ran Kayes card and my RAM. No Change.

Before all this, I had reformatted and striped using Intech's 3.1.1. The RAID volume refused to mount, so after several tries, I went back to SoftRAID 2.2.2.

I considered trying ExpressRAID 2.3.1, but it looked like a total pain in the ass and I dropped that idea.

I'm getting excellent transfer rates; 77 MB/s reads and 71 MB/s writes, but having to run with "Write-Through" checked, I'm taking a hit of 13% in CPU performance.

After all I've tried, I can only assume it is caused by either SoftRAID, Initio's firmware or both and I'm still not a happy camper.

Ton: what is the dash number of your logic board? Mine is -04 and Kaye's is -05, I think.

magician
12-05-2000, 03:59 AM
hmmm...Louie, I wonder if it could be even lower level than that? I'm thnking ROM revisions here...something that impacts the way the PCI controller works? See where I'm going with that?

Ton
12-05-2000, 04:05 AM
Louie, if you can tell me where to look for it I'll get back to you asap.
Btw. I'm following your efforts with great interest. Despite being SR the possible cause, from your experience it seems that there's no descent alternative. I'm running the raid for about 2 months now. At 67 / 61 read/writes its stable and fast as hell. Up to now I haven't experienced any file corruption either. Thanks for not giving up. http://www.macgurus.com/ubb/cool.gif
Ton

kaye
12-05-2000, 11:30 AM
Ton,

Bottom right side of the mobo, almost behind the square plastic that goes around the speaker cover. Mine says PCC 5000-0121-05, 05 being the revision level. On the back of my PTP under the screw just below the AC power connector is a tape which says AUG-1997, the build date I guess.

Louie,

That is disappointing. Well, the ram is eliminated as a possibility. I am running that same ram in the PTP and was running it in the picky Newer G3 S900 with no problems and it is 50ns as you say. I just was never able to find out the significance of the 3 instead of 2. Newer Tech told me it would work fine but, of course, never told me what the 3 meant, then you have to factor in whether they were just guessing or what.

We can also eliminate the possibility of your xlr8 card being the problem.

So we are down to SoftRAID or the Initio firmware or, as magician says, something low level. Maybe I should check the part numbers on my Bandit chips and whatever else to compare.

BTW, melt the xlr8, run it faster while you have it, and tell us where it stops. I never did try faster with the switches than 540/270 though it complained and did ask for incrementally faster, around 544/272 as I recall. I did accept those speedups. Running it faster, if it will, also might tell us something about your mobo. Don't worry about the card.

I guess you are to the point where trying ExpressRAID 2.3.1 might be worthwhile though it looked like a pain to me also. I hate to read the docs. k

Louie
12-05-2000, 03:54 PM
If ton's mobo dash number is 04 like mine, that will be at least one thing in common. If my memory serves me, it was the last version, -05, that caused the biggest problem for XLR8. They actually reverse-engineered it and modified the Carrier card to work properly on the PTP and all the rest.

Ton
12-06-2000, 09:26 AM
Louie, looked it up for you and my 9500 says NM3 94V -0. anyway thats all the print I could find on the mobo.
It's an early 9500/132 so that would be one of the first series I guess?
Doesn't seem likely that revionsnumbers of macs/clones mobo's are parallel would they?
Ton

Louie
12-06-2000, 06:23 PM
My mistake Ton. I was thinking you had a Power Tower Pro. All PTP's have a 9500 ROM, but Kaye doesn't have this Write-Through problem. I was looking for any commonality.

kaye
12-06-2000, 06:40 PM
Ton,

Not sure about Louie but I was thinking you have a PTP. Your configuration of drives and cards, dual Miles2 with a drive connected to each card?

Louie,

I just had a thought. HDT whether RAID or not, has in its driver (not mode page parameters), by default, an item checked called "Synchronize cache on shutdown". It says shutdown but it also does this on reboot. What I remember of this is that it saves what is in the drive's onboard cache before it allows a shutdown or reboot. This only applies to PCI machines. I encountered this on my NuBus Macs because there is about a 4 second delay before shutdown or reboot. I unchecked it since the HDT docs said it was only required for PCI machines.

What I wonder is whether SoftRAID, when you have dual Miles2, is properly flushing/saving what is in the cache of drives connected to both cards. It may only be properly handling the drive cache on one card. Just a thought. k

kaye
12-06-2000, 07:46 PM
Louie,

Just reread SoftRAID's site and saw this:

We have had a few odd reports which we have not been able to pin down to a specific cause with some G3 upgraded cards. What happens is that the system may not boot properly, or may have occasional or frequent hangs. If you have a G3 or G4 upgrade card, test your system with all SoftRAID driver acceleration disabled. If your system is stable only with SoftRAID driver acceleration disabled, then you may need to leave SoftRAID acceleration disabled on your system.

You might try disabling SoftRAID acceleration, disabling Write-Through, and try booting. k

Louie
12-06-2000, 07:57 PM
Hmmm. Can I do that without having to rebuild my RAID volume? I might try that and see if the trade-off is worse or better than Write-Through.

[This message has been edited by Louie (edited 06 December 2000).]

kaye
12-06-2000, 08:13 PM
Yes you can. Just start SoftRAID. In the menu bar select Driver (I'm on my P120 so this is from memory) which opens a window called Driver Parameters. Uncheck Write Acceleration and uncheck Read Acceleration and uncheck Photoshop Acceleration, if checked. Click OK. Done. See what happens.

If problem persists, open SoftRAID again. Now you want to play with reducing the number of Maximum Concurrent I/Os which will also reduce the Memory Pool Size. Click OK and see what happens.

I'm not saying that this is the way to run, just an experiment to see if the problem is SoftRAID. k

kaye
12-06-2000, 08:44 PM
Louie,

I went out to get the mail and realized I forgot to mention -

After launching SoftRAID, highlight one of the X15's, then go to the menu bar and select Driver...

After done, you must highlight the other X15, and make the driver changes for it. Sorry, k

Louie
12-06-2000, 09:21 PM
Right. I just tried turning off all SoftRAID acceleration, restarted, unchecked Write-Through and it froze. I ran ATTO with accelleration off and got essentially the same numbers; 77 and 71. Odd.

It was a good idea.

kaye
12-06-2000, 09:40 PM
Nuts! Oh well. BTW, turning all that stuff off didn't make a noticeable difference for me in ATTO but it did in MacBench 5. About 5% in Disk and almost 10% in Publishing Disk. I ran that test when I was building up the S900 and comparing the SoftRAID driver with the HDT 3.0.2 driver with 2x 10k Cheetahs striped on one Miles2. SoftRAID with all acceleration off was almost exactly equal to HDT 3.0.2 in MacBench 5. So the SoftRAID acceleration items use main memory to further tweak their driver for some performance gain. k

Ton
12-07-2000, 03:02 AM
Louie and Kaye:
My config. and issue in a nutshell:
pm 9500/Newertech G3 400/1MB/2 x Miles2/2 ext.10K cheetah's/2 internal 10k Cheetah's.
Only when the 2 miles/4 Cheetha's are bus duplexed to one SR volume I have the same problem as Louie: Without write thru enabled the machine freezes when the backside cache kicks in. Using the Miles independently makes it go away.

kaye
12-07-2000, 01:46 PM
Ton,

What OS are you running? And have you tried other/earlier/later versions? k

Ton
12-07-2000, 03:52 PM
Kaye,
I'm running 9.04 and trying another OS version didn't come to mind,
so maybe I will try this. I think the problem isn't only tied to the backside cache vs.bus but a combination of things. When I boot of another disk, I mean not the raid volume, I succeed in booting succesfully with enabled backside cache and "write back". It's just impossible to run apps from the raid. Either they will freeze almost instantly after startup or crash within a few minutes of use, especially when heavy disk access is required for instance while saving to disk or loading plug-ins or fonts. Dunno if Louie's experience regarding this is the same. I don't think it has anything to do with the upgraded processor though. I tried the original 604 and a slower 266 Mhz G3/512 kb backside cache with no change. Also I tried the latest drivers from Newertech/XLR8 and Sonnet for the G3 cards with no success so I really think it points to Softraid and/or the Miles firmware.
Ton

kaye
12-07-2000, 04:09 PM
Ton,

I think Louie's experience is the same but with more disastrous results, a couple of HD's, not part of the RAID, hosed.

So far on my PTP, with OS7.6.1 and 8.6 and 9.0.4, I have not duplicated this problem. I have yet to try OS8.0. Just got a replacement CD for 8.0. Loaned mine out and I think it is a goner. If I could duplicate it, hopefully without losing a drive, I could get permission to take it to Initio, but I have to take a machine with the problem staring us in the face.

I did suggest to Louie trying dual ATTO UL3Ds, running an X15 off of one channel on each card, and that I could send him my UL3D if he could borrow another. That way we could see if the problem is the Miles2 or firmware. Also suggested trying ATTO RAID to see if SoftRAID is the culprit. But, as you can understand, he is just a little jumpy about losing any more HD's. k

kaye
12-07-2000, 04:49 PM
I think I may have found another ATTO UL3D so if you are patient Ton, and Louie wants to try, I may be able to send him two UL3D's to see if it is an Initio problem. He already has the ATTO RAID I think in case SoftRAID is the culprit. Anyway, we may be able to get to the bottom of this. k

Louie
12-07-2000, 09:16 PM
Ton:

How did the problem manifest itself when you ran the 604e card? There was no backside cache to kick in.

Do you mean that it froze as soon as the XLR8 extension loaded?? (which it will do even w/o a G3 card). If so, this sounds like an XLR8 software problem.

[This message has been edited by Louie (edited 07 December 2000).]

Ton
12-08-2000, 04:07 AM
Louie:
I made a mistake in my post regarding the 604: It should have been: G3 266@133 mhz/512 k no go. 604 Ok. Sorry about the confusion I've caused.
Regarding the G3 drivers: I've used the Machspeed drivers and the Maxpowr drivers from Newer and I don't see much difference between the two. Bot drivers allow write thru setting and with both drivers I need those settings to boot or work reliably from the raid.

Louie and Kaye: I'm sorry to hear Louie lost a couple of drives in the process, I didn't know that. I have another disk in the 9500, an older IBM for emergency startups etc. and that was never affected by the problem. I just tried booting from this disk with "write back" enabled and it starts up fine. But it's not possible in this situation to run applications from the raid, they will all freeze.

As I've said before in a previous posting: I don't make a very big deal out of not being able to use the write back option. I love the speed of the raid and I'm willing to sacrifice a little processor performance for this. But in the meantime I'm following you efforts to get to the bottom of this with great interest.
Thank you

kaye
12-21-2000, 05:26 PM
I have installed the dual ATTO cards in my PTP with some interesting results. Go to the Step-byStep for PTP-250. k