Hang during backup to VXA drive (reproducible)

[dfrick]

This hang is 100% reproducible with Retrospect Desktop Backup 5.0

 

Release, but not under either of the Beta releases.

 

 

 

 

 

I am having a problem with Retrospect Desktop Backup 5.0 Release,

 

under Mac OS 10.1.3 on a PowerMac G4/Cube/450MHz/1GB/20GB+60GB+20GB

 

with an Ecrix VXA-1 SCSI tape drive connected via a BridgeIt FireWire-SCSI

 

adapter.

 

 

 

I used this configuration extensively under 4.3, Beta 1/Preview and

 

Beta 2 (5.0.153) and had no problems. I removed Beta 2 and installed

 

Retrospect 5.0 Release, and used the Default preferences (so that

 

it would start clean). I deleted the existing catalogs.

 

 

 

Here is the script: I am backing up several partitions (two on the

 

internal 20GB hard disk and an external 60GB FireWire drive) to

 

the tape drive; Recycle backup, all files, verification ON, hardware

 

compression ON.

 

 

 

Here is the problem: About 200MB into the backup, Retrospect stops

 

showing files being copied to the tape drive, and just sits with

 

a busy cursor. It appears to be a 'resource wait', i.e., it doesn't

 

crash or complain, and the cursor keeps spinning, but it's not

 

soaking up all of the CPU time. After a few seconds, the tape drive

 

goes to 'not writing data' (green), then finally comes to a stop. There

 

is no data going to the tape drive, and it isn't locked up nor does

 

it have an error status on the panel. It appears to be waiting for

 

instructions or data from Retrospect. Retrospect does not respond

 

to Pause or Stop, and I had to force-quit Retrospect to get out.

 

 

 

Then, I used the same script, but changed the destination to be an

 

external 20GB FireWire disk drive. Retrospect 5.0 Release successfully

 

backed up all of the volumes (until it filled the drive, several GB short

 

of completion). Thus, the problem isn't with the source volumes or

 

Retrospect 5 Release's ability to backup in general.

 

 

 

Here is what I did to try to narrow the problem: I did a complete

 

cold boot (including all devices), and tried it again. It stopped

 

on exactly the same file. I then switched the first volume in the backup

 

to be a different volume, in case there was something wrong with that

 

file. Retrospect again stopped, and it had copied about 200MB. I again did

 

a complete cold boot, and tried it again. It stopped at the same place

 

on the second volume. Thus, it's not something wrong with the volume

 

(I had also run Disk Repair, Disk First Aid, and Disk Warrior before

 

doing any of the backups, so I know both volumes were happy.)

 

I also made a fresh backup set and used a different tape; same problem.

 

 

 

Then I installed Beta 2 (5.0.153) on a spare volume, did a cold boot,

 

and successfully backed up both volumes and the external 60GB volume to

 

the tape drive, including verification passes, as I had done many times

 

during the beta test.

 

 

 

Thus, the problem was reproduced five times from cold boot, and is

 

not a problem with source volumes, catalogs, tape cartridge, tape drive,

 

cables or other hardware. It appears that the problem is within Retrospect

 

Desktop Backup, introduced between Beta 2 (5.0.153) and Release (5.0.201),

 

and involves the Ecrix VXA-1 driver.

 

 

 

 

 

I will be glad to provide further details and help if needed.

 

 

[ghafer]

I sympathize. I have almost the same configuration, including the Cube and VXA-1. I also had the hanging rewrite problem with Retrospect 5.0 but, oddly, it was the tape that turned out to be the culprit. I would have never guessed since 4.3 worked fine with the same tape. I'm on my third one--Ecrix replaced the tape--and I've been fine since then.

 

 

 

This is my third VXA=1 firewire drive, the other two replaced, under warranty, within one month. It does not give me much satisfaction to know my data are contained in such a fragile device.

 

 

 

-- Gary

[shadowspawn]

I would suggest trying another scsi-firewire bridge if you can, it may be the problem.

 

 

 

I am backing up on Mac OS X to a VXA via SCSI, and a VXA via firewire (first generation firewire vxa, uses the bridge supplied by Ecrix). I am having a few problems, but none that lead me to believe the tape unit is the issue.

[waveman]

You're describing the same problem I've been having (noted in several other posts). I've got a blue G3 with an Adaptec 2906 SCSI card and a VXA-1 drive. This combination has worked flawlessly for over a year with Retrospect 4.X (under OS 9) and Retrospect 5 beta (under OS X).

 

 

 

Every time I try to back up my partitioned 20GB drive, Retrospect hangs after around 180MB. As you note, the drive simply stops writing and Retrospect becomes unresponsive. The only way out is to force quit. At one point, the log file contained trouble communicating error entries, but no longer.

 

 

 

I've tried two different tapes, moving the SCSI card to a different slot, changing the SCSI cable and termintor, changing the SCSI ID, updating the drive's firmware, downgrading the Adaptec driver to the February 2002 version, and changing the order of partitions in the script, all to no avail.

 

 

 

Here's the interesting part: backups continue to work properly when I run Retrospect 5 under OS 9. Every time.

[sjnoland]

I am having a similar problem, only my system doesn't hang. I have a dual 500 Mhz G4 with a firewire VXA drive. Running OS 9.1. I have seen a pause while writing about every 200 MBytes, but the system doesn't seem to hang and will pick up the transfer after the tape drive goes through it's timeout, rewind-a-bit, and go ready sequence. The backups I have tried all completed OK, just took a lot longer that it would have had these pauses not bee there.

 

 

 

I have tried this with two separate file sets and two separate tapes (one brand new). This didn't happen on 4.3. Interestingly, I note that the hangs only happen during the copy phase - the read-back for the compare doesn't stop at all. This leads me to suspect that it may have something to do with the compression part of the write cycle.

 

 

 

I have sent an email to Dantz tech support, but have heard nothing back from them. Sounds like there is a real problem here that has different manifestations depending on OS.

 

 

 

 

[ghafer]

Correction: I now have the identical problem using the Ecrix VX-1 firewire and my Cube. I cannot backup now without having a hang up. Retrospect 5 is unusable.

 

 

 

--Gary

[waveman]

That's an interesting point... I've also noticed that Retrospect 5 pauses every 200MB or so. The tape rewinds a bit, the middle LED goes on, and then writing continues. This happens under OS 9 and OS X, and it does't slow down the process. But the OS 9 backups eventually complete, while the OS X backups stop cold. I never noticed this behavior with Retrospect 4.3.

 

 

 

Unless we all have faulty hardware and/or software, there's something wrong here.

 

 

 

Any comments from a Dantz representative?

[ghafer]

Today, a Dantz representative directed me to Ecrix's website for a firmware update; he described the update being bundled with VXA tool.

 

 

 

Unfortunately, neither turned out to be accurate. I have the latest firmware, which dates back to November of last year, and the VXA tool has no bundled firmware accompanying it. I checked the Exabyte site to find the same information. VersionTracker lists similar results.

 

 

 

I'm recataloging under 9.2.2 so that I can get back my old documents. Under X, the catalog became corrupted using Retrospect 5.

 

 

 

Am I having fun yet?

 

 

 

--Gary

[waveman]

I doubt it (the fun part, that is).

 

 

 

Updating the firmware in the VXA-1 made no difference for me. This evening, I saw a glimmer of hope when I read another post about disabling hardware and software compression, but that didn't help, either.

 

 

 

I realize that Dantz can't validate Retrospect on every CPU/backup device/interface combination, but the VXA-1 has become a mainstream backup solution, and I'd be interested to know if they've tried it with Retrospect 5 under OS X.

 

 

 

At least the tech support person from Ecrix has been very helpful. Although he doesn't have a solution, he's continuing to email suggestions. (I doubt that the problem is with the drive.)

 

 

[dthompsn]

I too have seen the hangs with my VXA-1 drive, OS X, and Retrospect 5. My VXA-1 drive is attached to a 2930U Adaptec SCSI card and after I saw the error communicating notes in the system log, I took everything else off the SCSI chain and tried with both the 1.02 adaptec drivers and 1.1 adaptec drivers. The 1.02 drivers show faster backup than the 1.1, so I'm wondering if it isn't something to do with the SCSI drivers. I tried to get Adaptec to tell me how to turn the debugging on or if I could get some help troubleshooting, but they said that my tech support time had run out--if I wanted to pay to help them troubleshoot this--well then.

 

 

 

I was just really glad to see others having this problem and that it isn't just me. I have gotten one good backup, but it took 4 times of rebooting the machine (once it locks, there's no killing it--that's another reason I think it may be a SCSI problem.) Anyway, hope to see a resolution soon.

 

 

[dthompsn]

So can anybody honestly say they've gotten this to work? I just did the update to 5.0.203 and now its even worse. I can't even really communicate with the device. It will say the drive is empty, but when I put in a tape, it just sits and grinds (running the little cleaner wheel) like it can't read the tape. How can I help get this fixed? I really need a good backup system and OS X and this just isn't cutting it.

[dthompsn]

So can anybody honestly say they've gotten this to work? I just did the update to 5.0.203 and now its even worse. I can't even really communicate with the device. It will say the drive is empty, but when I put in a tape, it just sits and grinds (running the little cleaner wheel) like it can't read the tape. How can I help get this fixed? I really need a good backup system and OS X and this just isn't cutting it.

 

 

 

After the drive stops and Retrospect hangs, I get the following message in system.log

 

Apr 10 13:36:19 localhost mach_kernel: Timeout: T/L = 4:0 Cmd = 02727c00 Cmd Type = 1

 

 

[dthompsn]

So can anybody honestly say they've gotten this to work? I just did the update to 5.0.203 and now its even worse. I can't even really communicate with the device. It will say the drive is empty, but when I put in a tape, it just sits and grinds (running the little cleaner wheel) like it can't read the tape. How can I help get this fixed? I really need a good backup system and OS X and this just isn't cutting it.

 

 

 

After the drive stops and Retrospect hangs, I get the following message in system.log

 

Apr 10 13:36:19 localhost mach_kernel: Timeout: T/L = 4:0 Cmd = 02727c00 Cmd Type = 1

 

 

 

I think Dantz better make the statement that SCSI VXA support is not available at this time.

[CallMeDave]

Six different issues, and almost no specific information provided by anyone!

 

 

 

The first poster, dfrick, was using a BridgeIt SCSI to FireWire adapter. No indication of the firmware on the adapter or the drive.

 

 

 

The second poster, ghafer, says he has "almost the same setup," but he doesn't provide the Forum with any specific information about his setup. Dantz says he needs a firmware upgrade, and he says the firmware of his VXA is up to date; what about the firmware of his BridgeIt adapter? Or is it some other adapter?

 

 

 

waveman, shadowspawn and dthompson have SCSI host adapters but don't provide details about which interface is on their VXA-1 (narrow or LVD); what sort of cables are being used? 50 to 25 pin adapters anywhere?

 

 

 

Waveman has a Blue & White G3; is it stock? No cpu upgrades (which have caused a problem for another user on the board, or at least the Xcelerate software was)?

 

 

 

Come on people, if we're going to help each other we have to be specific!

 

 

 

Or maybe I've just gotta get a new hobby...

 

 

 

Dave

[waveman]

Dave,

 

 

 

Actually, I have provided more specific information, but in other threads. In any case, my setup is as follows:

 

 

 

1. Blue G3 with a G4/500 OWC CPU card but no cache control software

 

 

 

2. Adaptec 2906 SCSI card

 

 

 

3. Cable with a 25 pin connector at the Mac end and a 50 pin "thin" connector at the VXA end

 

 

 

4. Most recent Adaptec OS X driver (1.1.0)

 

 

 

5. Most recent VXA firmware (2B7B)

 

 

 

I do agree that we ought to try to help one another, but at some point, someone from Dantz is going to have to step up to the plate and give us some ideas.

 

 

 

5. Mac OS 10.1.3

[IrenaS]

Those using the Bridgeit FireWire/SCSI converter, try updating the Firmware on the converter to the latest version, 302:

 

 

 

http://www.fastware.net/drivers/BIT_Downloads.html

 

 

 

For SCSI support, note that wide drives need to be connected to supported wide cards to be supported under OS X. If your set-up is qualified (by these conditions) and you are still having trouble, please post details of the card, the drive, the computer and relevant firmware.

 

 

 

We have actually tested the SCSI and FireWire VXA drives sucessfully in our labs, and have not seen these kinds of problems. If there is a commonality in the problems, of course we will want to look into it.

 

 

 

Thanks,

 

 

 

Irena Solomon

 

Dantz Tech Support

[waveman]

OK, here goes (again):

 

 

 

1. Blue G3 w G4/500MHz ZIF CPU card, no cache control software

 

 

 

2. Adaptec 2906 SCSI card

 

 

 

3. OS 10.1.3

 

 

 

4. Latest driver for SCSI card

 

 

 

5. VXA-1 external tape drive (SCSI 2) with 2B7B firmware. Drive passes hardware test and works fine under OS 9.

 

 

 

6. Retrospect 5 desktop, patched to latest version

 

 

 

Consistent backup failures during writing, usually after ~200MB. Retrospect stops writing to the drive and eventually must be force quit.

 

 

 

Thanks.

[CallMeDave]

One question:

 

 

 

Does it fail without the after-market third party hardware upgrade?

 

 

 

SCSI on OS X is iffy enough; that a Blue & White with a 500 Mhz G4 might not get the timing right doesn't sound too far fetched...

 

 

 

Dave

[waveman]

Dave,

 

 

 

I thought of that. Unfortunately, I can't test it without the upgrade - the G3 went back to OWC for a rebate (they say they never received it, but that's another story).

 

 

 

What strikes me is that the failures that other users are reporting are similar to mine, and I suspect that they don't all have CPU upgrades. There's got to be a common thread. Even if Retrospect is having trouble communicating with the drive because of SCSI flakiness under OS X, why does it hang and not report an error?

 

 

 

By the way, were you also experiencing problems with a VXA-1? I couldn't tell from your earlier post. If so, have you gotten anywhere?

 

 

 

Thanks.

 

 

 

-waveman

[jblarsen]

I just upgraded from the preview to the latest Retrospect and am seeing similar issues. I run a G4/733, OS 10.1.3, Adaptec 2930, Ecrix VXA drive.

 

 

 

Backup to the Ecrix is almost half the speed that it was with the Preview or OS 9. It was 180-210MB/min for a backup of a local disk, now it is 110-130MB/min.

 

 

 

I see a pause for about 4-5 seconds for every few hundred megabytes written.

 

 

 

No failures to write data yet, however I did not have verification on.

 

 

 

Also a few other errors like a network disconnect and assertion error... but not necessarily relevent to this thread.

[waveman]

Consider yourself lucky! I've also notice the pauses every 200MB or so... the drive stops writing, rewinds a little, stops, and then picks up again. When I run Retrospect under OS X, this process always breaks down at some point. Interestingly, I also get the pauses under OS 9, but the backup proceeds and finishes, albeit more slowly than 4.3 ever did. I've never tried turning verification off, but the hangs never occur during that phase.

[CallMeDave]

What strikes me is that the failures that other users are reporting are similar to mine, and I suspect that they don't all have CPU upgrades. There's got to be a common thread.

 

 

 

Seems that everyone having this problem is running SCSI under Mac OS X! Even the FireWire version of the VXA is a SCSI device with a bridge of dubious compatibility.

 

 

 

 

 

Even if Retrospect is having trouble communicating with the drive because of SCSI flakiness under OS X, why does it hang and not report an error?

 

 

I'd guess that Retrospect was designed to communicate with working SCSI, and may just not be equipped with the sort of error handling that is turning out to be necessary with SCSI under OS X.

 

 

 

I certainly don't see these sorts of problems doing File Type backups to IDE drives; they work every time...

 

 

 

Dave

 

(who hated SCSI under OS 9, and misses the agony now...)

[waveman]

I'd guess that Retrospect was designed to communicate with working SCSI, and may just not be equipped with the sort of error handling that is turning out to be necessary with SCSI under OS X.

 

 

 

I certainly don't see these sorts of problems doing File Type backups to IDE drives; they work every time...

 

 

 

Dave

 

(who hated SCSI under OS 9, and misses the agony now...)

 

 

 

I know what you mean... I always considered SCSI somewhat of a black art, but I usually managed to get recalcitrant chains to work, even back in the days when my Laserwriter was a SCSI device. I suspect that OS X has brought us to a whole new level.

 

 

 

The $64,000 question here is: "Who fixes this? Apple? Adaptec? Dantz?"

 

 

 

-waveman

[pendragon]

I'm also seeing the periodic stop, backup, stop and return to writing every 200 MB during backups. At first I thought the heads needed cleaning, but that was a waste of time.

 

 

 

G4/450 (768MB) - X Server 10.1.3 with all updates and Retro 5.0 Workgroup with latest patch for memory leak, ATTO ExpressPCI DC (1.6.6 firmware - latest) wide SCSI card, Ecrix VXA-1 wide SCSI (V22B7B firmware - latest). Tape drive shares one ATTO SCSI channel with an internal wide SCSI drive. Other ATTO SCSI channel has two wide SCSI drives.

 

 

 

SCSI setup seems bullet proof, never a hang, machine runs for months under heavy loads. Only rebooted to install software/firmware updates.

 

 

 

With same drive and card saw 180-200 MB/min on 4.3 and with 5 Preview. Same throughput with same tape drive and card on previous machine (SuperMac S900 250 MHz) running 9.1 and 4.3. Since 5 Workgroup was installed, I only get 105-115 MB/min because of the tape pausing. This is horrible.

 

 

 

Is Dantz now writing some kind of marker every 200 MB? I've also noticed that when each client/volume finishes backing up, the tape drive goes throught the same pause business, leaving the drive with the middle green light on. If the next client/volume starts right up, there is now quite a sizable delay before the tape starts streaming. On 4.3 and 5.0 Preview, the drive was left with the orange right light on and there was no delay when the next client started writing. When the backups were finished with the old software, the drive would eventually drop from the orange light to the middle green, but it took a number of minutes.

 

 

 

I never had any reliability or positioning problems with the old software, so I can hardly see why this change was necessary.

[seymour]

I'm having exactly the same problem. PowerMac G4/450 DP, Mac OS X 10.1.3, Adaptec 2906 SCSI (version

 

1.1 drivers) card and a VXA-1 SCSI drive (latest firmware).

 

 

 

I get about 2.5 GB into the backing up (after several short rewind, pause, resume iterations) and then the drive goes into the stopped state (middle light green) and says there. Trying the Retrospect 5.0.203 upgrade didn't help, either.

 

 

 

Since I've been waiting for a long time for Retrospect to work with my VXA driver under Mac OS X, it's extremely disappointing that the final release version doesn't work as well as the beta.