Jump to content

Hang during backup to VXA drive (reproducible)


Recommended Posts

  • Replies 73
  • Created
  • Last Reply

Thanks much for the insight you've provided on the state of the VXA-1 drive. Most of it correlates with my prior suspicions that the backspaces were related to a change in the tape handling as of the 5.0 release. You have identified that Dantz implemented a read check of some sort about every 200 MB when a file mark is written, to ensure data integrity across a power loss.

 

 

 

As you are aware this causes a performance penalty, which is very severe for those of us with fast buses and drives, which we bought so that backups of gigantic volumes could be completed in a timely fashion. In our case the penalty is almost 2X, and this places our entire Retrospect operation at risk, as we are very close to not being able to finish our incremental backups during off-hours when our users are not affected by back-up downtime. It is extremely important for us to guarantee that all our interrelated files are self-consistent in a snapshot. If users were allowed to run during a backup, after the backups overran their time slot, any restore we would do might not provide a consistent set of files, defeating the whole purpose. Shutting the users down is also a bad and costly option.

 

 

 

That said, we have UPS on our entire backup configuration, and most of our critical systems, so this data integrity problem can never occur for us. I would think other serious shops would do the same. The problem is that you are penalizing these shops that really need the performance and have paid for it 100% of the time. This is a questionable risk vs. benefit trade-off, considering that power failures are not a particularly common phenomenon, even without UPS.

 

 

 

Actually have written tape drivers in a past life, it confuses me a little as to what integrity is gained by reading a tape mark. If you want to ensure that your catalog is in sync with the tape, you could just update the catalog file marker count (or whatever) after you've transferred a full drive's cache of data following the write of a mark, or if you're conservative, you could update to the previous mark when you send the command to write the next.

 

 

 

The tape drive should balk if there is a media error during a write. Any user who doesn't trust the drive can check the verify box. If you have a power failure somewhere during the process, your catalog still knows the last guaranteed good mark location. Either scheme will provide the same integrity for no performance cost.

 

 

 

This might even improve the reliability and life of the VXAs and media because the mechanical stress will be significantly reduced by running the tape continuously. Since a lot of the VXA hangs with Adaptec/Firewire seem to correlate to the tape pause/200 MB boundaries, those folks might also see a lot better reliability, or they might actually be able to do backups again, like they did on 4.3 and 5.0 Preview. These are significant benefits.

 

 

 

At the very least, would Dantz consider a check-box preference for the backspaces so that UPS users and/or performance freaks could get back the speed we paid for and have depended on with our VXA drives and Retrospect for the past two years? Thanks again.

 

 

 

 

 

 

 

 

Link to comment
Share on other sites

True, but the bridge that Ecrix is using in internal, and it's always the same one. The reason that we can't support other stand-alone bridges is because we can't test every converter out there with every supported device. It's not the fact that there's a converter in the equation that makes it unsupported, it the combination of factors that potentially compromise the integrity of the backup just enough for us to say that we can't support it.

Link to comment
Share on other sites

dhwagner asked:

 

> I thought that the VXA firewire drive --at least early versions--was

 

> basically a SCSI drive with a firewire adapter. Is this incorrect?

 

 

 

You thought right, in fact the initial VXA firewire packages were sold with an external SCSI-firewire adaptor. I got one at a good price when the built-in firewire VXA drive shipped.

 

 

 

Thanks for the comments and links Irena. The company I work for produces data acquisition devices for use over SCSI and USB. We also decided not to officially support adaptors, there was just too much variation in behaviour.

Link to comment
Share on other sites

1. I too noticed the very noticible pausing with the VXA drive firewire model on Retrospect 5 (I have the latest rev, latest firmware, and os 10.1.4)

 

 

 

2. I also noticed that on completion of a script it doesn't rewind the tape and quit the software though I've configured it to do that..

 

 

 

 

Link to comment
Share on other sites

As stated, I have a VXA-1 external SCSI hooked to an Adaptec 2930U SCSI adapter. Both have the same pin interface and I use a cable of the correct connections at both ends. I still experience the hangs.

 

 

 

I did drop over to OS 9.2.2 to make sure that I haven't overlooked something in the drives configuration and I noticed in my hex dump logfile, the following line:

 

 

 

*** SCSI Log Page 37 Failed ***

 

 

 

p

Link to comment
Share on other sites

I have the same configuration as dthompsn and have the same problem. Three out of four attempts to backup stall when they first attempt to write to the drive and never complete. This didn't happen with either 4.3 or the 5.0 preview.

 

 

 

I can provide a full report Apple System Profiler if that would help anyone.

Link to comment
Share on other sites

Please provide details on your configuration:

 

 

 

Retrospect Version and build

 

Computer Model and speed

 

OS X version and build

 

Device (from Retrospect's Configure > Devices > Device Status window)

 

Device firmware (in the version column of the Device Status window)

 

Device interface

 

Media brand and size

 

Other Devices

 

Adapter Card

 

Adapter Card firmware

 

Adapter Card driver for OS X

 

 

 

Thanks,

 

 

 

Irena Solomon

 

Dantz Tech Support

 

 

Link to comment
Share on other sites

Retrospect Version and build: 5.0.205

 

Computer Model and speed: PowerPC G4 (AGP Graphics) 400 MHz

 

OS X version and build: Was 10.1.3 but now at 10.1.4 (5Q125) and still shows problem

 

Device: SCSI-A:4 ECRIX VXA-1 2B7B Ecrix VXA DC (4.08)

 

Device interface: SCSI-2

 

Media brand and size: All Ecrix, some V17 and some V10 all show same behavior

 

Other Devices: On this SCSI, I show within the device window,

 

SCSI-A:3 MATSHITA PD-2 LF-D100 A110 SCSI [My DVD-RAM drive].

 

Firewire-B MITSUMI CR-4804TE 2.4C Mitsumi CD-RW (1.50) [QPS Que!Fire].

 

 

 

As mentioned in an earlier reponse to this, I've tried disconnecting the DVD-RAM drive with same results. I also have on that SCSI interface an internal SCSI hard-drive.

 

 

 

Adapter Card: Adaptec 2930U

 

Adapter Card Firmware: 4.3

 

Adapter Card driver: Tried both 1.0.2 and 1.1 with same results.

Link to comment
Share on other sites

Retrospect Version and build: 5.0.205

 

Computer Model and speed: PowerPC G4 500MP

 

OS X version and build: Mac OS X 10.1.4 (5Q125)

 

Device: ECRIX VXA-1 Ecrix VXA DC (4.08)

 

Device firmware 2848

 

Device interface SCSI-A:3

 

Media brand and size : V17

 

Other Devices: No other SCSI devices

 

Adapter Card:

 

SCSI card

 

Card type: scsi

 

Card name: ADPT,2930CU

 

Card model: ADPT,1686806-04

 

Card ROM #: Not available

 

Card revision: 3

 

Card vendor ID: 9004

 

Adapter Card firmware: n/a

 

Adapter Card driver for OS X: 1.1.0

 

 

 

 

Link to comment
Share on other sites

Just to add to the collection I see the same stall after approx 200 MB with the following configuration:

 

 

 

Retrospect Version and build - workgroup 5.0.205

 

Computer Model and speed - B&W 350

 

OS X version and build - Server 10.1.4 5S40

 

Device (from Retrospect's Configure > Devices > Device Status window)

 

FireWire-A Ecrix VXA-1 2A6A Ecrix VXA DC (4.08)

 

Device firmware (in the version column of the Device Status window) - 2A6A

 

Device interface - FireWire

 

Media brand and size - VXA V17

 

Other Devices - NA

 

Adapter Card - motherboard FireWire

 

Adapter Card firmware - NA

 

Adapter Card driver for OS X - NA

 

 

Link to comment
Share on other sites

I absolutely agree!

 

 

 

Halving the backup speed to improve integrity in the face of rare-to-nonexistent failures

 

is a dubious tradeoff (and, it doubles the amount of time during which such a failure

 

could occur).

 

 

 

Please, please, give us the option to run at full speed!

 

 

Link to comment
Share on other sites

  • 2 weeks later...

This thread seems to have been quiet for a few days.

 

 

 

Does this mean that the problem has gone away for everyone else, or is everyone (like me) stuck with a non-working backup mechanism?

 

 

 

Is there a way to reactivate the v5 preview - this worked better than the upgrade I paid $199 for.

Link to comment
Share on other sites

Is there a way to reactivate the v5 preview - this worked better than the upgrade I paid $199 for.

 

 

 

You mean the one you didn't pay anything for?

 

 

 

A similar thread on the RetroTALK email list described information from Exebyte saying that Dantz has addressed the issue for a future update (an RDU) (hmmmm, list archives appear to be offline; thread was called "New %&^\@ Backup System")

 

 

 

Seems that they really _did_ work with the drive's manufacturer to provide both better performance _and_ data integrity.

 

 

 

Probably better use of their time then re-enabling an unsupported beta...

 

 

 

Dave

 

 

 

 

 

 

Link to comment
Share on other sites

I've seen these comments. However, they don't address my situation as I am trying to use an Ecrix supplied external FireWire VXA drive. The only parties involved are

 

* Apple (G3 350) who supply the host FireWire interface,

 

* Ecrix/Exabyte who have integrated the tape drive and FireWire

 

* Dantz

 

 

 

This all worked with V5 preview - but does not work now.

Link to comment
Share on other sites

Looking at your posts, it seems that the problem you are experiencing is the pause in writing every 200 MB and, thus, slower performance. Is this correct? If so, this was explained previously in this thread, and is also in the Knowledgebase. It's something that will be addressed, and we will post information here when it's available.

 

 

 

If you are seeing other problems, please post details.

 

 

 

Thanks,

 

 

 

Irena Solomon

 

Dantz Tech Support

Link to comment
Share on other sites

Well, that's not really a response to me, since initially posted on 5/1 and followed up with the requested configuration information on 5/2. Since then, no reponse from from Dantz, not even that they are looking into the problem. As a result, I can't backup, with a setup that worked fine under Mac OS 9 and under Mac OS X with the preview.

 

 

 

Still waiting for a response.

Link to comment
Share on other sites

In reply to:

rtsuk writes:

 

initially posted on 5/1 and followed up with the requested configuration information on 5/2. Since then, no reponse from from Dantz, not even that they are looking into the problem.


 

Are you kidding? They've said that yes, a change was made to the vxa driver between beta and release and are working with the manufacturer of the drive to provide speed and reliability. And Ecrix is apparently telling its customers that Dantz is preparing to release an update to address the issue.

 

 

 

Since the users who are getting hangs do so at the 200Mb stalls, I'd guess that once they fix the latter the former will go away too.

 

 

 

Dave

Link to comment
Share on other sites

I think what I am seeing is an interaction between used tapes and the new "pause every 200MB" feature.

 

 

 

With 5 tapes (1 new, 4 used) I had:

 

- the new one tape worked first time

 

- 2 used tapes worked first time

 

- 1 used tape stalled at 200 MB and then stopped for ever

 

but was usable after a long erase

 

- 1 used tape stalled at 200 MB and then stopped for ever

 

and did the same after the first long erase but

 

was OK after a second long erase

 

 

 

Short erases didn't seem to help. If it gets past the first 200 MB pause it seems to keep going until the end.

 

 

 

All these were full backups. I've not yet tried an incremental or a recycle backup. So, I don't know if the problems will come back.

 

 

 

Nick

 

 

Link to comment
Share on other sites

  • 2 weeks later...

Greetings,

 

 

 

Retrospect Driver Update (RDU) 2.6 contains new drivers for Exabyte/Ecrix VXA tape drives. This RDU is only for Retrospect or Retrospect 5.0 for Mac. It may be downloaded now with the link below, but the Dantz website will not be updated until next week.

 

 

 

ftp://ftp.dantz.com/pub/updates/rdu26.sit

 

 

 

Thanks.

 

 

 

Irena Solomon

 

Dantz Development

Link to comment
Share on other sites

  • 2 weeks later...

It's back!

 

 

 

New Retrospect Driver Updates for current Mac and Windows versions of Retrospect are now available for download from the Dantz website.

 

 

 

Mac RDU 2.7 is recommended for all Retrospect 5.0 users. In addition to new drivers for a handful of previously supported devices, RDU 2.7 adds support for the following drives:

 

 

 

Benchmark VS640

 

Exabyte VXA AutoPak 1x10

 

EZQuest Boa CDR/w 24X10X40 FireWire Ext

 

HP SureStore Tape Autoloader 1/8 Series

 

VST SmartDisk FWCDRW-V8

 

 

 

It also includes improved handling of the following drives:

 

 

 

Exabyte/Ecrix VXA tape drives

 

Que! Firewire 8x4x32x CD drive (prevents the assert at cdr.c-6249)

 

 

 

The RDU may be downloaded from:

 

 

 

http://www.dantz.com/index.php3?SCREEN=rdu

 

 

 

Thanks,

 

 

 

Irena Solomon

 

Dantz Tech Support

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...